U.S. patent application number 12/193581 was filed with the patent office on 2009-10-01 for user adjustable volume control that accommodates hearing.
Invention is credited to Ronald D. Blum, William R. Saunders, Michael A. Vaudrey.
Application Number | 20090245539 12/193581 |
Document ID | / |
Family ID | 27490051 |
Filed Date | 2009-10-01 |
United States Patent
Application |
20090245539 |
Kind Code |
A1 |
Vaudrey; Michael A. ; et
al. |
October 1, 2009 |
USER ADJUSTABLE VOLUME CONTROL THAT ACCOMMODATES HEARING
Abstract
A method for processing audio signals optimizes the listening
experience for hearing impaired individuals to feel stigmatized by
requiring them to employ special hearing-impaired equipment. A user
actuated controller controls a mixture of a preferred audio signal
and a remaining audio signal across a range sufficiently wide
enough to encompass all individuals. The preferred audio is
recorded and maintained separate from all remaining audio and
delivered to the listener in a manner that maintains the
separateness of the preferred audio and the remaining audio. The
user actuated controller includes the capability of automatically
maintaining the listener established ratio in the face of changes
in the audio signal. The user actuated controller enables the user
to specify a range about the ratio in which the audio may vary,
which permits the listener to expand the audio across a continuous
range to whatever dynamic range his hearing can accommodate. The
controller automatically adjusts to changes in incoming audio. The
controller can react to relatively slowly moving changes or prevent
short bursts of sound in the remaining audio from modifying the
signal levels. The combination of the above aspects provides a
heretofore not possible listening experience that can accommodate
the listening desires of all listeners. The combination of the
ability to control the ratio of preferred audio to remaining audio
and to specify the dynamic range about the ratio in which the audio
may vary, coupled with the ability of the controller to
automatically adjust the signal levels in response to sudden
changes in incoming audio, provides a powerful user capability that
truly optimizes the listening experience for any listener.
Inventors: |
Vaudrey; Michael A.;
(Blacksburg, VA) ; Saunders; William R.;
(Blacksburg, VA) ; Blum; Ronald D.; (Roanoke,
VA) |
Correspondence
Address: |
PERKINS COIE LLP;PATENT-SEA
P.O. BOX 1247
SEATTLE
WA
98111-1247
US
|
Family ID: |
27490051 |
Appl. No.: |
12/193581 |
Filed: |
August 18, 2008 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
09673275 |
Oct 13, 2000 |
7415120 |
|
|
12193581 |
|
|
|
|
09059307 |
Apr 14, 1998 |
|
|
|
09673275 |
|
|
|
|
09059303 |
Apr 14, 1998 |
|
|
|
09059307 |
|
|
|
|
09059304 |
Apr 14, 1998 |
|
|
|
09059303 |
|
|
|
|
Current U.S.
Class: |
381/109 |
Current CPC
Class: |
H03G 7/002 20130101;
H04R 5/04 20130101; H04S 3/00 20130101; H04R 2205/041 20130101;
H04S 2420/03 20130101 |
Class at
Publication: |
381/109 |
International
Class: |
H03G 3/02 20060101
H03G003/02 |
Claims
1-35. (canceled)
36. An audio system for optimizing playing of an audio program,
which includes a preferred audio signal and a remaining audio
signal, for end users, which includes both hearing impaired and
non-hearing impaired listeners, comprising: a first end user
adjustable amplifier that receives and amplifies the preferred
audio signal to a level specified by the user; a second end user
adjustable amplifier that receives and amplifies the remaining
audio signal to a level specified by the user; a first soft
clipping circuit receiving as an input the preferred audio signal,
having an output coupled to the input of the first end user
adjustable amplifier, and limiting an overall magnitude of the
preferred audio signal when a level of the preferred audio signal
exceeds a first predetermined value; and a second soft clipping
circuit receiving as an input the remaining audio signal, having an
output coupled to the input of the second end user adjustable
amplifier, and limiting an overall magnitude of the remaining audio
signal when a level of the remaining audio signal exceeds a second
predetermined value, the combination of the first and second soft
clipping circuits and first and second end user adjustable
amplifiers thereby automatically maintaining a user selected ratio
of the preferred audio signal to remaining audio signal, even in
the presence of transient changes in either the preferred audio
signal or the remaining audio signal.
37. The system according to claim 36, wherein the first and second
predetermined values in the first and second soft clipping
circuits, respectively, are user definable values.
38-51. (canceled)
52. The system according to claim 36, wherein the preferred audio
signal comprises a voice signal.
53. The system according to claim 36, wherein the preferred audio
signal comprises a non-voice signal.
54. The system according to claim 36, wherein the remaining audio
signal comprises a background signal.
55. The system of claim 36, wherein the first and second soft
clipping circuits output normalized versions of their inputs based
on a preset parameter.
56. An audio system for optimizing playing of an audio program,
which includes a preferred audio signal and a remaining audio
signal, for end users, which includes both hearing impaired and
non-hearing impaired listeners, comprising: first adjustable
amplifier means under end user control that receives and amplifies
the preferred audio signal to a level specified by the user; second
adjustable amplifier means under end user control that receives and
amplifies the remaining audio signal to a level specified by the
user; first circuit means for receiving as an input the preferred
audio signal, having an output coupled to the input of the first
adjustable amplifier means, and limiting an overall magnitude of
the preferred audio signal when a level of the preferred audio
signal exceeds a first predetermined value; and second circuit
means for receiving as an input the remaining audio signal, having
an output coupled to the input of the second adjustable amplifier
means, and limiting an overall magnitude of the remaining audio
signal when a level of the remaining audio signal exceeds a second
predetermined value, the combination of the first and second
circuit means and first and second adjustable amplifier means
thereby automatically maintaining a user selected ratio of the
preferred audio signal to remaining audio signal, even in the
presence of transient changes in either the preferred audio signal
or the remaining audio signal.
57. The system according to claim 36, wherein the first and second
predetermined values in the first and second circuit means,
respectively, are user definable values.
58. The system according to claim 36, wherein the preferred audio
signal comprises a voice signal.
59. The system according to claim 36, wherein the preferred audio
signal comprises a non-voice signal.
60. The system according to claim 36, wherein the remaining audio
signal comprises a background signal.
61. The system of claim 36, wherein the first and second circuit
means output normalized versions of their inputs based on a preset
parameter.
Description
RELATED APPLICATIONS
[0001] The present application is related to U.S. patent
application Ser. No. 08/852,239 entitled "Hearing Enhancement
System and Method" filed on May 6, 1997 by Ronald D. Blum and
William E. Kokonaski (HEC-1), which is hereby incorporated by
reference, including the drawings, as if repeated herein in its
entirety.
[0002] The present invention is also related to U.S. patent
application Ser. No. 08/907,503 entitled "Listening Enhancement
System and Method" filed on Aug. 8, 1997 by Ronald D. Blum and
William E. Kokonaski (HEC-2), which is hereby incorporated by
reference, including the drawings, as if repeated herein in its
entirety.
[0003] The present application is a continuation-in-part of U.S.
patent application Ser. No. 09/059,307 entitled "Improved Hearing
Enhancement System and Method" filed on Apr. 14, 1998 by Ronald D.
Blum, William E. Kokonaski, William R. Saunders and Michael A.
Vaudrey (HEC-3), which is hereby incorporated by reference,
including the drawings, as if repeated herein in its entirety.
[0004] The present application is also a continuation-in-part of
U.S. patent application Ser. No. 09/059,303 entitled "Improved
Listening Enhancement System and Method" filed on Apr. 14, 1998 by
Ronald D. Blum, William E. Kokonaski, William R. Saunders and
Michael A. Vaudrey (HEC-4), which is hereby incorporated by
reference, including the drawings, as if repeated herein in its
entirety.
[0005] The present application is also a continuation-in-part of
U.S. patent application Ser. No. ______ entitled "Improved Aural
System and Method" filed on Apr. 14, 1998 by William R. Saunders
and Michael A. Vaudrey (HEC-5), which is hereby incorporated by
reference, including the drawings, as if repeated herein in its
entirety.
[0006] The present application also claims priority to U.S.
Provisional Application No. 60/109,506 entitled "Special
Application Digital Audio Decoder" filed on Nov. 23, 1998 by
William R. Saunders and Michael A. Vaudrey (HEC-6), which is hereby
incorporated by reference, including the drawings, as if repeated
herein in its entirety.
BACKGROUND OF THE INVENTION
[0007] The present invention relates generally to methods and
apparatus for processing audio, and more particularly to a method
and apparatus for processing audio to improve the listening
experience for a broad range of listeners, including hearing
impaired listeners.
[0008] As one ages and progresses through life, over time due to
many factors, such as age, genetics, disease, and environmental
effects, one's hearing becomes compromised. Usually, the
deterioration is specific to certain frequency ranges.
[0009] In addition to permanent hearing impairments, one may
experience temporary hearing impairments due to exposure to
particular high sound levels. For example, after target shooting or
attending a rock concert one may have temporary hearing impairments
that improve somewhat, but over time may accumulate to a permanent
hearing impairment. Even lower sound levels than these but longer
lasting may have temporary impacts on one's hearing, such as
working in a factory or teaching in a elementary school.
[0010] Typically, one compensates for hearing loss or impairment by
increasing the volume of the audio. But, this simply increases the
volume of all audible frequencies in the total signal. The
resulting increase in total signal volume will provide little or no
improvement in speech intelligibility, particularly for those whose
hearing impairment is frequency dependent.
[0011] While hearing impairment increases generally with age, many
hearing impaired individuals refuse to admit that they are hard of
hearing, and therefore avoid the use of devices that may improve
the quality of their hearing. While many elderly people begin
wearing glasses as they age, a significantly smaller number of
these individuals wear hearing aids, despite the significant
advances in the reduction of the size of hearing aids. This
phenomenon is indicative of the apparent societal stigma associated
with hearing aids and/or hearing impairments. Consequently, it is
desirable to provide a technique for improving the listening
experience of a hearing impaired listener in a way that avoids the
apparent associated societal stigma.
[0012] Most audio programming, be it television audio, movie audio,
or music can be divided into two distinct components: the
foreground and the background. In general, the foreground sounds
are the ones intended to capture the audiences attention and retain
their focus, whereas the background sounds are supporting, but not
of primary interest to the audience. One example of this can be
seen in television programming for a "sitcom," in which the main
character's voices deliver and develop the plot of the story while
sound effects, audience laughter, and music fill the gaps.
[0013] Currently, the listening audience for all types of audio
media are restricted to the mixture decided upon by the audio
engineer during production. The audio engineer will mix all other
background noise components with the foreground sounds at levels
that the audio engineer prefers, or at which the audio engineer
understands have some historical basis. This mixture is then sent
to the end user as either a single (mono) signal or in some cases
as a stereo (left and right) signal, without any means for
adjusting the foreground to the background.
[0014] The lack of this ability to adjust foreground relative to
background sounds is particularly difficult for the hearing
impaired. In many cases, programming is difficult to understand (at
best) due to background audio masking the foreground signals.
[0015] There are many new digital audio formats available. Some of
these have attempted to provide capability for the hearing
impaired. For example, Dolby Digital, also referred to as AC-3 (or
Audio Codec version 3), is a compression technique for digital
audio that packs more data into a smaller space. The future of
digital audio is in spatial positioning, which is accomplished by
providing 5.1 separate audio channels: Center, Left and Right, and
Left and Right Surround. The sixth channel, referred to as the 0.1
channel refers to a limited bandwidth low frequency effects (LFE)
channel that is mostly non-directional due to its low frequencies.
Since there are 5.1 audio channels to transmit, compression is
necessary to ensure that both video and audio stay within certain
bandwidth constraints. These constraints (imposed by the FCC) are
more strict for terrestrial transmission than for DVD, currently.
There is more than enough space on a DVD to provide the user with
uncompressed audio (much more desirable from a listening
standpoint). Video data is compressed most commonly through MPEG
(moving pictures experts group) developed techniques, although they
also have an audio compression technique very similar to
Dolby's.
[0016] The DVD industry has adopted Dolby Digital (DD) as its
compression technique of choice. Most DVD's are produced using DD.
The ATSC (Advanced Television Standards Committee) has also chosen
AC-3 as its audio compression scheme for American digital TV. This
has spread to many other countries around the world. This means
that production studios (movie and television) must encode their
audio in DD for broadcast or recording.
[0017] There are many features, in addition to the strict encoding
and decoding scheme, that are frequently discussed in conjunction
with Dolby Digital. Some of these features are part of DD and some
are not. Along with the compressed bitstream, DD sends information
about the bitstream called metadata, or "data about the data." It
is basically zero's and ones indicating the existence of options
available to the end user. Three of these options that are relevant
to HEC are dialnorm (dialog normalization), dynrng (dynamic range),
and bsmod (bit stream mode that controls the main and associated
audio services). The first two are an integral part of DD already,
since many decoders handle these variables, giving end users the
ability to adjust them. The third bit of information, bsmod, is
described in detail in ATSC document A/54 (not a Dolby publication)
but also exists as part of the DD bitstream. The value of bsmod
alerts the decoder about the nature of the incoming audio service,
including the presence of any associated audio service. At this
time, no known manufacturers are utilizing this parameter. Multiple
language DVD performances are provided via multiple complete main
audio programs on one of the eight available audio tracks on the
DVD.
[0018] The dialnorm parameter is designed to allow the listener to
normalize all audio programs relative to a constant voice level.
Between channels and between program and commercial, overall audio
levels fluctuate wildly. In the future, producers will be asked to
insert the dialnorm parameter which indicates the level (SPL) at
which the dialog has been recorded. If this value is set as 80 dB
for a program but 90 dB for a commercial, the television will
decode that information examine the level the user has entered as
desirable (say 85 dB) and will adjust the movie up 5 dB and the
commercial down 5 dB. This is a total volume level adjustment that
is based on what the producer enters as the dialnorm bit value.
[0019] A section from the AC-3 description (from document A/52)
provides the best description of this technology. "The dynrng
values typically indicate gain reduction during the loudest signal
passages, and gain increase during the quiet passages. For the
listener, it is desirable to bring the loudest sounds down in level
towards the dialog level, and the quiet sounds up in level, again
towards dialog level. Sounds which are at the same loudness as the
normal spoken dialogue will typically not have their gain
changed."
[0020] The dynrng variable provides the user with an adjustable
parameter that will control the amount of compression occurring on
the total volume with respect to the dialog level. This essentially
limits the dynamic range of the total audio program about the mean
dialog level. This does not, however, provide any way to adjust the
dialog level independently of the remaining audio level.
[0021] One attempt to improve the listening experience of hearing
impaired listeners is provided for in The ATSC, Digital Television
Standard (Annex B). Section 6 of Annex B of the ATSC standard
describes the main audio services and the associated audio
services. An AC-3 elementary stream contains the encoded
representation of a single audio service. Multiple audio services
are provided by multiple elementary streams. Each elementary stream
is conveyed by the transport multiplex with a unique PID. There are
a number of audio service types which may be individually coded
into each elementary stream. One of the audio service types is
called the complete main audio service (CM). The CM type of main
audio service contains a complete audio program (complete with
dialogue, music and effects). The CM service may contain from 1 to
5.1 audio channels. The CM service may be further enhanced by means
of the other services. Another audio service type is the hearing
impaired service (HI). The HI associated service typically contains
only dialogue which is intended to be reproduced simultaneously
with the CM service. In this case, the HI service is a single audio
channel. As stated therein, this dialogue may be processed for
improved intelligibility by hearing impaired listeners.
Simultaneous reproduction of both the CM and HI services allows the
hearing impaired listener to hear a mix of the CM and HI services
in order to emphasize the dialogue while still providing some music
and effects. Besides providing the HI service as a single dialogue
channel, the HI service may be provided as a complete program mix
containing music, effects, and dialogue with enhanced
intelligibility. In this case, the service may be coded using any
number of channels (up to 5.1). While this service may improve the
listening experience for some hearing impaired individuals, it
certainly will not for those who do not employ the proscribed
receiver for fear of being stigmatized as hearing impaired.
Finally, any processing of the dialogue for hearing impaired
individuals prevents the use of this channel in creating an audio
program for non-hearing individuals. Moreover, the relationship
between the HI service and the CM service set forth in Annex B
remains undefined with respect to the relative signal levels of
each used to create a channel for the hearing impaired.
[0022] Other techniques have been employed to attempt to improve
the intelligibility of audio. For example, U.S. Pat. No. 4,024,344
discloses a method of creating a "center channel" for dialogue in
cinema sound. This technique disclosed therein correlates left and
right stereophonic channels and adjusts the gain on either the
combined and/or the separate left or right channel depending on the
degree of correlation between the left and right channel. The
assumption being that the strong correlation between the left and
right channels indicates the presence of dialogue. The center
channel, which is the filtered summation of the left and right
channels, is amplified or attenuated depending on the degree of
correlation between the left and right channels. The problem with
this approach is that it does not discriminate between meaningful
dialogue and simple correlated sound, nor does it address unwanted
voice information within the voice band. Therefore, it cannot
improve the intelligibility of all audio for all hearing impaired
individuals.
[0023] The separation of voice from background audio in television
signals is discussed by Shiraki in U.S. Pat. No. 5,197,100. The
technique employed by Shiraki involves the use of band pass
filtering in combination with summing and subtracting circuits to
form a "voice channel" that would be differentiated from the rest
of the audio programming. The limitation of this approach is that
the band pass filter only discriminates frequencies within a
predetermined range, in this case 200 Hz to 500 Hz. It cannot
discriminate between voice and background audio that may happen to
fall within the band pass frequency. Furthermore, the application
of band pass filtering cannot distinguish between relevant and
irrelevant speech-components within an audio signal.
[0024] Means of reducing background noise in audio frames have been
discussed by Solve et al. in U.S. Pat. No. 5,485,522 which
discloses a speech detector and a noise estimator used to
adaptively adjust attenuation to each frame of an audio signal.
This and other forms of Adaptive Noise Filtering cannot distinguish
between voice and other non-stationary audio in the voice band,
such as music or irrelevant voice information. Consequently, the
improvement in intelligibility is less than optimum.
[0025] Attempts to improve the listening experience by modifying
the signal level in the face of large noise variations have been
made. For example, U.S. Pat. No. 5,434,922 to Miller et al.
discloses a method and system for sound optimization which measures
both the music and noise in a vehicle. Miller et al. uses analog to
digital conversions and adaptive filtering with algorithms to
compensate for the ambient noise background by enhancing the sound
signal automatically. This technique cannot compensate for a
preferred audio signal that is overwhelmed by the remaining audio
signal for a particular listener. In the system of Miller et al.,
the system merely increases the total signal level in an attempt to
overcome the presence of road and engine noise. In most cases,
audio that was not intelligible to a particular listener does not
become intelligible by merely increasing the signal level.
[0026] In general, prior art techniques employing band pass
filtering or selective equalization will not remove voice band
background or noise within the voice band range from speech
components of the audio program. The previously cited inventions of
Dolby, Shiraki and Miller et al. have all attempted to modify some
content of the audio signal through various signal processing
hardware or algorithms, but those methods do not satisfy the
individual needs or preferences of different listeners. In sum, all
of these techniques provide a less than optimum listening
experience for hearing impaired individuals as well as non-hearing
impaired individuals.
[0027] Finally, in the case of studio recordings, vocals are
usually recorded separately and are later mixed with the
instrumentals and placed on a single recording track. The end user
is therefore enabled to only adjust the volume, tone and balance
(in the case of stereo), but not the relative signal levels of the
voice component or the background component.
[0028] The present invention is therefore directed to the problem
of developing a system and method for processing audio signals that
optimizes the listening experience for hearing impaired listeners,
as well as non-hearing impaired listeners, without forcing hearing
impaired individuals to feel stigmatized by requiring them to
employ special hearing-impaired equipment.
SUMMARY OF THE INVENTION
[0029] The present invention solves this problem by providing a
user actuated controller that controls a mixture of a preferred
audio signal and a remaining audio signal across a range
sufficiently wide enough to encompass all individuals, from hearing
impaired listeners to non-hearing impaired listeners. By providing
a controller that all listeners employ to optimizes their listening
experience to their unique hearing requirements, the present
invention prevents the attachment of any societal stigma to the use
of such a controller. Moreover, the present invention therefore
does not require the use of specialized equipment for the hearing
impaired, and the concomitant costs usually associated with such
special equipment due to the small manufacturing quantities, and
the higher percentage of research and development recovery costs
associated with each unit.
[0030] According to another aspect of the present invention, the
preferred audio is recorded and maintained separate from all
remaining audio and delivered to the listener (or end user) in a
manner that maintains the separateness of the preferred audio and
the remaining audio. This enables the end user to have complete
control over the ratio of the mixture, which can be adjusted by the
end user to accommodate his or her hearing impairments at the time
of playing the audio, rather than at the time the audio is being
recorded. Moreover, maintaining the two audio sources separate
permits the user to make adjustments depending upon the type of
audio being played, the composition of the audience, i.e., to
attempt to find a compromise setting that meet the needs of several
individuals listening to the audio, the exact surroundings,
including external noise sources, room acoustics, etc. In fact, the
present invention thus provides unlimited flexibility in playing
the audio that was heretofore not possible.
[0031] According to another aspect of the present invention, the
user actuated controller includes the capability of automatically
maintaining the listener established ratio in the face of changes
in the audio signal. In other words, if the signal level of the
remaining audio suddenly changes, the controller automatically
adjusts the signal level of either the preferred audio or the
signal level of the remaining audio to compensate. In this case, if
the action moves to an outdoor windy area that tends to obscure the
preferred audio, the listener need not re-adjust the ratio setting
to compensate. Rather, the controller automatically corrects for
this change.
[0032] According to yet another aspect of the present invention,
the user actuated controller enables the user to specify a range
about the ratio in which the audio may vary to provide for a user
specifiable dynamic range. This permits the listener to expand the
audio across a continuous range to whatever dynamic range his
hearing can accommodate. So, a young listener may set a value for
the preferred audio to remaining audio ratio and then open the
dynamic range completely, thereby providing for complete freedom of
the audio, whereas an older listener with compromised hearing may
specify a very narrow range about the specified ratio, in which
range the audio is limited to vary. Reducing the dynamic range to
the point where the preferred audio signal is maintained without
variation creates a monotone signal, which may be necessary for a
severely hearing impaired individual to understand the audio, but
this may not be required for an individual with a milder hearing
impairment. The adjustability of this range lets the user create
his optimum listening experience, which may vary greatly depending
upon multiple factors that can be best accounted for at the time of
playing the audio program rather than at the time of recording the
audio program.
[0033] According to yet another aspect of the present invention,
the controller automatically adjusts to changes in incoming audio,
however, the controller can react to relatively slowly moving
changes or prevent short bursts of sound in the remaining audio
from modifying the signal levels. Thus, a gun shot may or may not
overwhelm any preferred audio depending on the user's listening
desires.
[0034] The combination of the above aspects of the present
invention provides a heretofore not possible listening experience
that can accommodate the listening desires of all listeners. The
combination of the ability to control the ratio of preferred audio
to remaining audio and to specify the dynamic range about the ratio
in which the audio may vary, coupled with the ability of the
controller to automatically adjust the signal levels in response to
sudden changes in incoming audio, provides a powerful user
capability that truly optimizes the listening experience for any
listener.
BRIEF DESCRIPTION OF THE DRAWINGS
[0035] FIG. 1 illustrates a general approach according to the
present invention for separating relevant voice information from
general background audio in a recorded or broadcast program.
[0036] FIG. 2 illustrates an exemplary embodiment according to the
present invention for receiving and playing back the encoded
program signals.
[0037] FIG. 3 illustrates one exemplary technique in which the
preferred signal can be delivered to the end user through voice
recognition.
[0038] FIG. 4 shows a circuit diagram of a digital version of the
voice recognition device depicted in FIG. 3.
[0039] FIG. 5 illustrates a generalized block diagram of a path
that a preferred signal versus a remaining audio signal might take
for any general media type before it arrives at the end user
according to the present invention.
[0040] FIG. 6 illustrates a prior art block diagram of the signal
paths taken by the audio and video components of a television
program before arriving at the end user.
[0041] FIG. 7 illustrates an exemplary embodiment according to the
present invention that incorporates the ability to transmit two
separate audio signals simultaneously.
[0042] FIG. 8 illustrates the use of a unique stereo mixture from
the production process to deliver a relatively pure preferred
signal content to the end user without affecting the normal
stereophonic audio programming, which can be applied to any media
in general, that utilizes stereo programming.
[0043] FIG. 9 illustrates an exemplary gain adjustment embodiment
according to the present invention for the preferred signal versus
remaining audio.
[0044] FIG. 10 illustrates inclusion of an equalization process
into a gain adjustment to create a frequency dependent gain
according to one aspect of the present invention.
[0045] FIG. 11 illustrates a simplified block diagram of an
exemplary embodiment according to the present invention for
soft-clipping of both the preferred signal and remaining audio for
the purpose of automatically maintaining a user selected preferred
signal to remaining audio (PSRA) ratio.
[0046] FIG. 12 illustrates an exemplary embodiment according to the
present invention for implementing a soft-clipping method as shown
in FIG. 11.
[0047] FIG. 13 is a detailed block diagram illustrating an
exemplary method according to the present invention for maintaining
the level of either the preferred audio signal or the remaining
audio signal in the presence of transients in either signal.
[0048] FIG. 14 illustrates the first step in performing ambient
noise measurement for the purpose of automatic volume adjustment,
system identification according to one aspect of the present
invention.
[0049] FIG. 15 is a block diagram illustrating the automatic volume
adjustment process to compensate for changes in the ambient noise
in a listening environment according to another aspect of the
present invention.
[0050] FIG. 16 illustrates the concept according to the present
invention of controlling the companding in terms of the amplitudes
of the time trace signals for voice and remaining audio, termed the
dynamic amplitude range (DAR).
[0051] FIG. 17 shows an approximate moving average of voice (solid)
and background audio (dotted), in which the average volume of the
remaining audio frequently exceeds the average volume of the voice,
making dialog difficult to understand.
[0052] FIG. 18 illustrates the effect on an adjustment on the
signals depicted in FIG. 17, even though the VRA has been corrected
during some segments of the program, other sections of the
remaining audio are still too loud due to transients.
[0053] FIG. 19 shows the sections where such an auto-leveling
process would work on the same signals depicted in FIGS. 17-18, in
which it is shown how some average DAR is retained in the
signals.
[0054] FIG. 20 represents the way audio is currently designed to be
transmitted to the end user on DVD using the dialog mode and the
multi-language services.
[0055] FIG. 21 depicts the decoder components that perform the
adding, as altered according to the present invention.
[0056] FIG. 22 illustrates the intended spatial positioning setup
of a common home theater system.
[0057] FIG. 23 illustrates a system where the user has the option
to select the automatic VRA leveling feature or the calibrated
audio feature according to the present invention.
[0058] FIG. 24 illustrates one conceptual diagram of how a downmix
would be implemented according to the present invention.
[0059] FIG. 25 shows that this downmix capability according to the
present invention could be built into the primary audio decoder
that is utilized by any DVD, DTV, or any other digital audio
playback device that is intended to decode and distribute audio
signals to the speakers.
[0060] FIG. 26 depicts a Dolby digital prior art encoder and
decoder.
[0061] FIG. 27 illustrates the user adjustable levels on each of
the decoded 5.1 channels according to the present invention.
[0062] FIG. 28 illustrates the interface box depicted in FIG. 27,
according to the present invention.
[0063] FIG. 29 depicts the results of two experiments conducted on
a random sample of the population including elementary school
children, middle school children, middle-aged citizens and senior
citizens, measuring the voice-to-remaining audio setting preferred
by each subject.
DETAILED DESCRIPTION
[0064] While much of the following description includes examples of
the technologies used in conjunction with digital Dolby, the
present invention can be used with any audio codec with multiple
transmission capability, such as DTS, THX, SDDS, PCM, etc.
Significance of Ratio of Preferred Audio to Remaining Audio
[0065] The present invention begins with the realization that the
listening preferential range of a ratio of a preferred audio signal
relative to any remaining audio is rather large, and certainly
larger than ever expected. This significant discovery is the result
of a test of a large sample of the population regarding their
preferences of the ratio of the preferred audio signal level to a
signal level of all remaining audio.
Specific Adjustment of Desired Range for Hearing Impaired or Normal
Listeners
[0066] Very directed research has been conducted in the area of
understanding how normal and hearing impaired users perceive the
ratio between dialog and remaining audio for different types of
audio programming. It has been found that the population varies
widely in the range of adjustment desired between voice and
remaining audio.
[0067] Two experiments have been conducted on a random sample of
the population including elementary school children, middle school
children, middle-aged citizens and senior citizens. A total of 71
people were tested. The test consisted of asking the user to adjust
the level of voice and the level of remaining audio for a football
game (where the remaining audio was the crowd noise) and a popular
song (where the remaining audio was the music). A metric called the
VRA (voice to remaining audio) ratio was formed by dividing the
linear value of the volume of the dialog or voice by the linear
value of the volume of the remaining audio for each selection.
These ratios are plotted in FIG. 29. For the music the minimum
ratio was 0.1 (indicating very little voice) while the maximum
ratio was 9.54 (indicated a large amount of voice). The sports test
had a minimum ratio of 0.33 and a maximum ratio of 20.17.
[0068] Several things were made clear as a result of this testing.
First, no two people prefer the identical ratio for voice and
remaining audio for both the sports and music media. This is very
important since the population has relied upon producers to provide
a VRA (which cannot be adjusted by the consumer) that will appeal
to everyone. This can clearly not occur, given the results of these
tests. Second, while the VRA is typically higher for those with
hearing impairments (to improve intelligibility) those people with
normal hearing also prefer different ratios than are currently
provided by the producers.
[0069] It is also important to highlight the fact that any device
that provides adjustment of the VRA must provide at least as much
adjustment capability as is inferred from these tests in order for
it to satisfy a significant segment of the population. Since the
video and home theater medium supplies a variety of programming, we
should consider that the ratio should extend from at least the
lowest measured ratio for any media (music or sports) to the
highest ratio from music or sports. This would be 0.1 to 20.17, or
a range in decibels of 46 dB. It should also be noted that this is
merely a sampling of the population and that the adjustment
capability should theoretically be infinite since it is very likely
that one person may prefer no crowd noise when viewing a sports
broadcast and that another person would prefer no announcement.
Note that this type of study and the specific desire for widely
varying VRA ratios has not been reported or discussed in the
literature or prior art.
[0070] In this test, an older group of men was selected and asked
to do an adjustment (which test was later performed on a group of
students) between a fixed background noise and the voice of an
announcer, in which only the latter could be varied and the former
was set at 6.00. The results with the older group were as
follows:
TABLE-US-00001 TABLE I Individual Setting 1 7.50 2 4.50 3 4.00 4
7.50 5 3.00 6 7.00 7 6.50 8 7.75 9 5.50 10 7.00 11 5.00
[0071] To further illustrate the fact that people of all ages have
different hearing needs and preferences, a group of 21 college
students was selected to listen to a mixture of voice and
background and to select, by making one adjustment to the voice
level, the ratio of the voice to the background. The background
noise, in this case crowd noise at a football game, was fixed at a
setting of six (6.00) and the students were allowed to adjust the
volume of the announcers' play by play voice which had been
recorded separately and was pure voice or mostly pure voice. In
other words, the students were selected to do the same test the
group of older men did. Students were selected so as to minimize
hearing infirmities caused by age. The students were all in their
late teens or early twenties. The results were as follows:
TABLE-US-00002 TABLE II Student Setting of Voice 1 4.75 2 3.75 3
4.25 4 4.50 5 5.20 6 5.75 7 4.25 8 6.70 9 3.25 10 6.00 11 5.00 12
5.25 13 3.00 14 4.25 15 3.25 16 3.00 17 6.00 18 2.00 19 4.00 20
5.50 21 6.00
[0072] The ages of the older group (as seen in Table I) ranged from
36 to 59 with the preponderance of the individuals being in the 40
or 50 year old group. As is indicated by the test results, the
average setting tended to be reasonably high indicating some loss
of hearing across the board. The range again varied from 3.00 to
7.75, a spread of 4.75 which confirmed the findings of the range of
variance in people's preferred listening ratio of voice to
background or any preferred signal to remaining audio (PSRA). The
overall span for the volume setting for both groups of subjects
ranged from 2.0 to 7.75. These levels represent the actual values
on the volume adjustment mechanism used to perform this experiment.
They provide an indication of the range of signal to noise values
(when compared to the "noise" level 6.0) that may be desirable from
different users.
[0073] To gain a better understanding of how this relates to
relative loudness variations chosen by different users, consider
that the non-linear volume control variation from 2.0 to 7.75
represents an increase of 20 dB or ten (10) times. Thus, for even
this small sampling of the population and single type of audio
programming it was found that different listeners do prefer quite
drastically different levels of "preferred signal" with respect to
"remaining audio." This preference cuts across age groups showing
that it is consistent with individual preference and basic hearing
abilities, which was heretofore totally unexpected.
[0074] As the test results show, the range that students (as seen
in Table II) without hearing infirmities caused by age selected
varied considerably from a low setting of 2.00 to a high of 6.70, a
spread of 4.70 or almost one half of the total range of from 1 to
10. The test is illustrative of how the "one size fits all"
mentality of most recorded and broadcast audio signals falls far
short of giving the individual listener the ability to adjust the
mix to suit his or her own preferences and hearing needs. Again,
the students had a wide spread in their settings as did the older
group demonstrating the individual differences in preferences and
hearing needs. One result of this test is that hearing preferences
is widely disparate.
[0075] Further testing has confirmed this result over a larger
sample group. Moreover, the results vary depending upon the type of
audio. For example, as shown in FIG. 29, when the audio source was
music, the ratio of voice to remaining audio varied from
approximately zero to about 10, whereas when the audio source was
sports programming, the same ratio varied between approximately
zero and about 20. In addition, the standard deviation increased by
a factor of almost three, while the mean increased by more than
twice that of music.
[0076] The end result of the above testing is that if one selects a
preferred audio to remaining audio ratio and fixes that forever,
one has most likely created an audio program that is less than
desirable for a significant fraction of the population. And, as
stated above, the optimum ratio may be both a short-term and
long-term time varying function. Consequently, complete control
over this preferred audio to remaining audio ratio is desirable to
satisfy the listening needs of "normal" or non-hearing impaired
listeners. Moreover, providing the end user with the ultimate
control over this ratio allows the end user to optimize his or her
listening experience.
[0077] The end-user's independent adjustment of the preferred audio
signal and the remaining audio signal will be the apparent
manifestation of one aspect of the present invention. To illustrate
the details of the present invention, consider the application
where the preferred audio signal is the relevant voice
information.
Creation of the Preferred Audio Signal and the Remaining Audio
Signal
[0078] FIG. 1 illustrates a general approach to separating relevant
voice information from general background audio in a recorded or
broadcast program. There will first need to be a determination made
by the programming director as to the definition of relevant voice.
An actor, group of actors, or commentators must be identified as
the relevant speakers.
[0079] Once the relevant speakers are identified, their voices will
be picked up by the voice microphone 1. The voice microphone 1 will
need to be either a close talking microphone (in the case of
commentators) or a highly directional shot gun microphone used in
sound recording. In addition to being highly directional, these
microphones 1 will need to be voice-band limited, preferably from
200-5000 Hz. The combination of directionality and band pass
filtering minimize the background noise acoustically coupled to the
relevant voice information upon recording. In the case of certain
types of programming, the need to prevent acoustic coupling can be
avoided by recording relevant voice of dialogue off-line and
dubbing the dialogue where appropriate with the video portion of
the program. The background microphones 2 should be fairly
broadband to provide the full audio quality of background
information, such as music.
[0080] A camera 3 will be used to provide the video portion of the
program. The audio signals (voice and relevant voice) will be
encoded with the video signal at the encoder 4. In general, the
audio signal is usually separated from the video signal by simply
modulating it with a different carrier frequency. Since most
broadcasts are now in stereo, one way to encode the relevant voice
information with the background is to multiplex the relevant voice
information on the separate stereo channels in much the same way
left front and right front channels are added to two channel stereo
to produce a quadraphonic disc recording. Although this would
create the need for additional broadcast bandwidth, for recorded
media this would not present a problem, as long as the audio
circuitry in the video disc or tape player is designed to
demodulate the relevant voice information.
[0081] Once the signals are encoded, by whatever means deemed
appropriate, the encoded signals are sent out for broadcast by
broadcast system 5 over antenna 13, or recorded on to tape or disc
by recording system 6. In case of recorded audio video information,
the background and voice information could be simply placed on
separate recording tracks.
Receiving and Demodulating the Preferred Audio Signal and the
Remaining Audio
[0082] FIG. 2 illustrates an exemplary embodiment for receiving and
playing back the encoded program signals. A receiver system 7
demodulates the main carrier frequency from the encoded audio/video
signals, in the case of broadcast information. In the case of
recorded media, the heads from a VCR or the laser reader from a CD
player 8 would produce the encoded audio/video signals.
[0083] In either case, these signals would be sent to a decoding
system 9. The decoder 9 would separate the signals into video,
voice audio, and background audio using standard decoding
techniques such as envelope detection in combination with frequency
or time division demodulation. The background audio signal is sent
to a separate variable gain amplifier 10, that the listener can
adjust to his or her preference. The voice signal is sent to a
variable gain amplifier 11, that can be adjusted by the listener to
his or her particular needs, as discussed above.
[0084] The two adjusted signals are summed by a unity gain summing
amplifier 12 to produce the final audio output. In this manner the
listener can adjust relevant voice to background levels to optimize
the audio program to his or her unique listening requirements at
the time of playing the audio program. As each time the same
listener plays the same audio, the ratio setting may need to change
due to changes in the listener's hearing, the setting remains
infinitely adjustable to accommodate this flexibility.
Voice Recognition Used to Create Audio Channel
[0085] Referring to FIG. 3, there is shown a diagram of yet another
exemplary embodiment of the present invention, which is a system
100 utilizing a voice recognition chip 101 that separates the
information into a mostly voice channel 102 and a background noise
channel 103 after recognizing the speech components of the incoming
signal. The mostly voice-like components are separated from the
background components and grouped into separate data streams. The
signals are then converted back to analog signals to create a
mostly voice channel and a background noise channel.
[0086] As in FIG. 2, the mostly voice channel 102 is fed to a
signal level adjustable amplifier 104, which performs the voice
volume adjustment, and the background channel 103 is fed to a
second signal level adjustable amplifier 105, which performs the
background volume adjustment. The outputs of the amplifiers 104,
105 are summed in a summing amplifier 106 to create the final
audio.
Voice Recognition (Digital Version)
[0087] FIG. 4 shows a circuit diagram of the digital version of the
voice recognition device 101. The audio signal is received,
converted into a digital signal and separated into digital voice
108 and digital background 109 signals by voice recognizer 107. The
digital signals are then converted back to an analog signal by
digital-to-analog converters 110, 111.
Voice Recognition Chip Tailored to User's Malady
[0088] A further embodiment of the instant invention utilizes a
chip which is programmable and by initialization is tailored to
each user's malady. For example, the user's particular frequency
loss can be separated out from the incoming signal, and passed
through the mostly voice channel, thereby enabling the user to
control the ratio of these frequencies relative to all other
frequencies. It is understood that the system and method are
adaptable to both analog and digital signals.
[0089] For the present invention, if the application is to increase
the listening pleasure of the end-user, it is possible to change
the voice microphone 1 to read the preferred signal microphone and
the background microphone 2 to read the remaining audio signal
microphone. Then, the rest of the means for providing the
end-listener independent adjustment of the preferred signal and
remaining audio will follow exactly as the preceding
discussion.
[0090] The present invention provides the end user with the ability
to adjust the level of the foreground audio (also called the
preferred signal) relative to the background audio (also called the
remaining audio). In the most general sense, the preferred signal
represents any audio component which is the immediate focus in the
programming while the remaining audio represents all other sound
components delivered simultaneously with that signal. It should be
noted that the remaining audio is still considered to be important
programming for the enjoyment of the total audio signal, but in its
current design, may be obscuring more important components in the
total audio signal. For example, an orchestral movement may contain
a rather lengthy flute solo which is the immediate focus. (Many
hearing impaired persons have the most trouble with the upper
frequency range, frequencies where the flute sounds mainly reside).
In most cases it would not be desirable to completely eliminate all
instruments except for the flute, even though the flute is the
primary focus of this particular movement. The remaining orchestral
instruments add to the enjoyment of the flute solo, but in order to
hear it more clearly, the end user needs to adjust the level of the
flute only to his/her own taste and hearing needs.
[0091] The idea of adjusting the preferred signal volume level over
the remaining audio level in a total audio signal spans all current
and future media. By way of example and not as an exhaustive list,
the present invention is applicable to television, CD, DVD, video
tape, movies, computer animation, and radio, audio transmitted over
the Internet, flash RAM storage, and computer generated audio.
Overall Process of the Present Invention
[0092] FIG. 5 illustrates an exemplary embodiment of the overall
process according to the present invention for creating the
required signal for any media format in general and for delivering
that required signal to the end user, which in this case is the
ordinary consumer or purchaser. Essentially, there are four main
steps in this process. Step 1 (element 51), the program to be
played to the end user is developed. Step 2 (element 52), the
preferred audio is obtained separately from the remaining audio.
Step 3 (element 53), the preferred signal and remaining audio are
delivered to the end user in a manner that maintains the
separateness of the preferred signal and remaining audio. Step 4
(element 54), the end user is provided with the ability to adjust
the relative signal levels of the preferred signal and remaining
audio.
Alternative Embodiment
[0093] For each of these mediums, there exist ways to obtain (step
2) and deliver the preferred signal to the end user (step 3), with
minimal impact on the listener who is satisfied with the current
preferred signal to remaining audio (PSRA) mix, determined by an
audio engineer. However, it has been shown that individuals with
generally normal hearing prefer to have control over the PSRA
mixture for at least some forms of programming.
[0094] One aspect of the present invention described herein begins
with possible ways for certain media formats, in which the
preferred signal can be delivered to the end user without affecting
the current mixture of the total audio signal. Then, turning toward
the end user adjustment controls (step 4), without loss of
generality, several additional aspects of the present invention are
presented.
Separate Broadband Equalization of the Preferred Audio Signal and
the Remaining Audio Signal
[0095] First, improvements in the flexibility of adjusting the
preferred signal to remaining audio mixture are realized in the
form of broadband equalization of the preferred and remaining audio
signals separately. This will provide the end user with even more
flexibility in reaching his/her desired listening quality above and
beyond that which is achievable using simple gain adjustment.
Maintaining the PSRA Ratio in the Face of Large Transients
[0096] Next, an aspect of the present invention is discussed in
which the preferred signal to remaining audio (PSRA) adjustment
selected by the end user and can be stored and maintained in the
presence of transients, which may occur in the background audio
also residing in the total audio signal. Several ways in which this
can be accomplished are explained in detail.
Adjustment of the Background Noise to Control the PSRA Ratio
[0097] Finally, a new invention which monitors the noise in the
environment in which the end user is residing will permit the total
audio programming, the preferred signal, or the remaining audio
signal, to be automatically adjusted in response to increases in
ambient room noise. This will prevent the end user from having to
continually adjust the volume of the audio programming when ambient
noise begins to mask the preferred signal.
Delivery of Preferred Signal Separate from the Remaining Audio
[0098] As mentioned above, the first segment of these embodiments
focuses on step 3 in FIG. 5. Delivering the preferred signal
separately from the remaining audio may vary drastically for each
media type. Several techniques are disclosed herein for delivering
the preferred signal to the end user in a format which it can be
separately adjusted with respect to the remaining audio, if
desired. It should be noted that each of these embodiments
initially assumes that the preferred signal is made available to
the delivery format during production of the audio program.
[0099] In general, audio programming usually begins with a separate
measurement of the preferred signal and is later condensed to a
mono or stereo signal before the end user receives it. Many of
these embodiments assume that the preferred signal is obtained
(step 2) early in the production process, but are not limited to
other possible methods for obtaining the preferred signal in
post-production.
Prior Art Recording and Delivery Process
[0100] FIG. 6 illustrates a very simplified block diagram of the
signal paths taken by the audio and video components of television
programming. At some point during the production process 65, the
total audio signal originates as a separate measurement of the
preferred and the remaining audio signals. Continuing with the
"sitcom" example described earlier, the preferred signal is likely
to be the actor's voices where all other signals such as music and
sound effects are remaining audio.
[0101] Early in the production process these signals are mixed by
mixer 66 under control of the audio engineer for delivery over a
portion of the total aural carrier 67 for a specific television
broadcast channel, either over airwaves or cable. This total audio
signal along with the video signal is received by the end user at
the television 68 where a single gain adjustment mechanism 69
allows the end user to adjust only the volume of the total audio
signal. This represents a very simplified version of the prior art
process of delivering an audio signal for a television program.
Recording and Delivery Process According to the Present
Invention
[0102] FIG. 7 illustrates an exemplary embodiment of a process
according to the present invention for delivering the preferred
signal and the remaining audio in a way that permits the end user
to adjust the respective signal levels. In this exemplary
embodiment, the mixing of the two audio signals (preferred signal
and remaining audio) is not done at the production studio, but
rather by the end user on his/her receiver 73. During production,
the audio engineer does not combine the preferred signal and
remaining audio, but rather sends the two audio signals separately
(element 70) on different parts of the total aural carrier (which
is created in element 71) in a portion of the bandwidth allocated
for each channel to transmit audio. This audio signal is
transmitted sent to the television receiver 73, in which the audio
signals are demodulated (element 72), sent through separate user
adjustable gain components and finally summed 74 to form the total
audio signal. This configuration allows the end user to adjust a
separated preferred signal component's volume with respect to the
remaining audio.
Restoration of the Normal Mix
[0103] In the alternative, provisions are made to restore the
normal mix for the end users who are currently satisfied with the
PSRA ratio. These are described next along with a more detailed
description of the total aural carrier in an analog scheme.
[0104] The total aural carrier is a 120 kHz bandwidth transmitted
on every television channel, intended to contain audio programming
which is played in conjunction with the video programming on that
channel. Currently, the most common components of the total aural
carrier are the mono and stereo segments allocated from baseband to
3f.sub.H respectively, where f.sub.H is 15.734 kHz. In addition,
centered around 5f.sub.H is the secondary audio programming (SAP)
channel, which is slightly lower in bandwidth than the mono and
stereo channels. Most modern day televisions and VCR's come
equipped with an audio decoder that will selected any one of these
three audio programs. While some programming is broadcast in
stereo, thus utilizing both the mono (50 Hz-12 kHz) and stereo (50
Hz-12 kHz) bandwidths, many television networks currently do not
transmit any information on the SAP channel (50 Hz-8 kHz). There
are also other segments of the total aural carrier that are not
utilized or televisions are not currently equipped with the
hardware to decode that segment of the bandwidth. Centered around
6.5f.sub.H is the low bandwidth "professional channel" and the
segments from 3 f.sub.H to 4f.sub.H and just above and below the
professional channel are not specifically allocated.
[0105] What this represents in terms of the preferred signal
transmission, is that there exists significant unutilized bandwidth
in conventional television transmission standards that can
accommodate additional signals transmitted in conjunction with the
normal audio program.
[0106] One exemplary embodiment of the present invention is to send
the remaining audio alone in the mono bandwidth while the preferred
signal is transmitted in the stereo bandwidth. This is not
particularly desirable because it will make the transmission of
stereo programming impossible.
[0107] Another alternative which retains the stereo signal is to
transmit the normal audio stereo programming as originally intended
but to transmit the preferred signal on another segment of the
total aural carrier, such as the SAP channel. This would require
that the television receiver be equipped with a way to
simultaneously decode both the stereo and the SAP channels, permit
separate gain adjustment on each channel, and add them together.
For normal audio programming the user needs only to select the
option that mutes the SAP channel or does not decode it at all. The
preferred signal audio can also be transmitted on any other segment
of the total aural carrier without deviating from the intended
invention.
[0108] As described above, the transmission of the preferred signal
simultaneously with the remaining audio or the total audio program,
can allow the end user to adjust the two separately to provide
improved listening enjoyment. This cannot be accomplished however,
without additional hardware that is not standard in current
televisions or VCR's.
[0109] For the case where the preferred audio signal is carried on
a bandwidth separate from the mono and stereo channels, a more
sophisticated demodulator will be required before the gain circuits
are applied. Current demodulators will deliver the mono, stereo, or
SAP channels separately (where stereo is generated from the
subtraction and addition of the mono and stereo channels with
appropriate division by two).
Additional Embodiment
[0110] There is one additional method in which the preferred signal
can be delivered to the end user that will provide normal stereo
programming for the population, requires only the stereo and mono
transmission bandwidths, and still allows relatively pure preferred
signal to be delivered to the end user. Since this method can also
be performed in all other stereo based media such as CD's, audio
tapes, VHS tapes, and radio, it is described in a general sense
with possible application to all stereo based media.
[0111] FIG. 8 illustrates one possible way in which the preferred
signal can be delivered to the end user utilizing only the stereo
left and right channels while still preserving the stereo image.
This method still relies on the production method (75) to alter the
original stereo mix for delivery of the preferred signal. There are
also several assumptions which are made that are generally, but not
always true; and are required to be satisfied for this method to
work to its full potential. The signals originating from the
production 75 of any audio programming are assumed to originate
from several different components, one of which is a pure or mostly
pure measurement of the preferred signal. (This assumption was also
made earlier for the description of television transmission methods
for the preferred signal). Another assumption which was inherently
made earlier but not explicitly stated was that the preferred
signal resides in center audio, that is equally distributed on the
left and right channels.
[0112] In any of the exemplary embodiments, the voice can easily be
panned to either the left or right stereo channel, but the original
desires (for the panning of the preferred signal) of the audio
engineers and audio production personnel will be lost with these
inventions. Finally, the most critical assumption being made for
this particular embodiment is that the remaining audio stereo image
is predominantly developed outside of the preferred signal
bandwidth. For example, if the preferred signal happens to be voice
and the remaining audio is special sound effects and music, the
frequencies from approximately 200 Hz to 4 kHz (voice bandwidth) of
the remaining audio signal are predominantly centered audio and the
remaining audio outside those bandwidths can create the stereo
image. As long as these conditions are met, the following technique
will provide a pure or nearly pure preferred signal to the end user
without compromising the stereo image of the normal audio
programming.
[0113] FIG. 8 begins with the production process 75 where the two
stereo signals are delivered to the end user as in FIG. 6. The
primary difference here is that the audio engineer pans the
preferred signal (P) to either the full right or full left (shown)
channel. Then each channel contains equal components of the
centered audio (CA) while the left channel also contains strictly
left audio (LA) and the right channel contains strictly right audio
(RA). After these two signals have been delivered to the end user
as a stereo program (CD, audio tape, VHS, television, etc.),
several steps follow in order to deliver both the normal and
adjusted audio mixtures.
[0114] First, each of the produced left and right stereo signals
are filtered with bandpass filters 76 intended to remove all
content that has not been identified as being in the band
containing the preferred signal. Again, for voice this will likely
be 200 Hz to 4 kHz. Following the previous assumption, this
filtering process will also remove components of both left and
right audio channels that are strictly left and right audio, namely
RA and LA, leaving only CA and P on the left (shown) or right
channel and CA on the right (shown) or left channel. Upon
subtraction (adder 77) of these two signals, a separate measurement
of the preferred audio signal P remains.
[0115] To regain the normal audio program with the preferred audio
signal centered, the separate measurement of the preferred audio
signal simply needs to be added (at adder 79) back to the right
audio signal. The left audio signal will remain unchanged from the
original delivery.
[0116] To accomplish a user adjusted audio mixture, the preferred
signal from the output of adder 77 is modified by a user adjustable
gain (W) before being added (adder 78) to the left and right normal
audio programming to form a centered preferred signal which can be
modified by the end user. The adjusted audio 80 is thus available
as the outputs of adders 78, whereas the normal or unadjusted audio
81 is available at the output of adder 79 and the original left
audio channel.
Less than Infinite Panning
[0117] With the method described above, several assumptions were
emphasized. One of these assumptions can be accounted for and
relaxed with some modification to the above process. Information
about the stereophonic location of the preferred signal can be
retained if the amount of panning of the preferred signal is a
known value that is less than infinite. In the above discussion,
the preferred signal was panned 100% to one channel during the
production process, allowing the filtered subtraction to produce a
nearly pure preferred signal. If it is known and established early
in the recording and delivery process that the preferred signal
will be panned to one channel by some finite ratio of left to right
(or right to left), then the stereo image of the preferred signal
can be reconstructed in the delivery process.
[0118] For example, suppose the desired original programming has a
preferred signal with an image that is 20 dB on the left channel
and 18 dB on the right channel. If a standard is established that
instructs audio engineers to pan the preferred signal to the left
exactly 10 dB, the new image is 30 dB on the right and 8 dB on the
left. Subtracting these new channels results in a 22 dB preferred
voice signal. This is then automatically adjusted to a level of 10
dB, a certain voltage level that can be easily referenced if such a
standard is put into practice. Now, the 10 dB preferred signal is
subtracted from the left channel and added to the right to restore
the original desired stereo image. A similar process is used to
create the same stereo image for the adjusted audio mixture.
[0119] The above can be applied to any media utilizing stereophonic
sound which originates from multiple measurements of the
programming, one including a relative pure measurement of the
preferred signal. These media may include CD's, VHS tapes, radio,
and television to name a few.
[0120] Having covered some components for delivery of the preferred
signal for certain audio media, several more generalized aspects of
the present invention are described next which can be used on any
platform supporting the preferred signal to remaining audio user
adjusted mixture. There are three separate aspects of the present
invention that are described which provide the end user with the
ability have more control over the PSRA mix, the behavior of the
total audio programming in the face of transients, and the audio
programming with respect to ambient noise in the listening
environment. Without loss of generality, each of the aspects of the
present invention can be coupled together and used in conjunction
with one another as well as with any of the delivery methods
described above. In fact, as long as a preferred signal and
remaining audio exists, any of the following exemplary embodiments
can be implemented to provide improvement in listening quality
above and beyond simple volume adjustment.
Frequency Dependent Gain Adjustment to PSRA
[0121] One of the original goals of providing the end user with a
preferred signal audio which is separate but in conjunction with
the remaining audio program, was to allow the hearing impaired to
improve their understanding of speech or other preferred signal
programming. While adjusting the overall volume level of the
preferred signal with respect to the volume level of the remaining
audio will improve intelligibility and enjoyment for the hearing
impaired (as well as the normal hearing population), it is not
sufficient for many types of hearing impairment. One aspect of the
present invention seeks to improve upon simple gain adjustment of
the preferred signal to remaining audio by introducing a frequency
dependent gain mechanism that is user controlled to amplify or
suppress certain frequency components of either the preferred
signal or remaining audio.
[0122] FIG. 9 illustrates an exemplary embodiment utilizing only
straight gain mechanisms. The preferred signal (PS) is passed
through a user adjustable gain circuit 92 and added by adder 94 to
the remaining audio signal which has been amplified by a separately
adjustable user gain element 93. Undoubtedly this format will
provide the end user with more flexibility than is currently
available in standard audio electronics.
[0123] However, the hearing impaired listener (who stands to
benefit from this technology) may be somewhat disappointed in the
apparent improvements. Most hearing impairments occur first in the
high frequency ranges, typically above 2 kHz. Taking, for example,
speech as the preferred signal; speech intelligibility is most
affected by the speech frequencies above 2 kHz. If the hearing
impairment of the listener resembles a low pass filter (as many do)
simply raising the gain of the speech may not improve speech
intelligibility because the low frequencies are now so loud that
they are not enjoyable.
[0124] The most effective way to compensate for this limitation in
gain adjustment, is frequency dependent gain adjustment. If the end
user could specify the frequency region where added volume is
required, the sound could be tailored much more specifically to the
hearing impairment.
[0125] FIG. 10 illustrates the inclusion of such a frequency
dependent gain device 96 on the preferred signal and on the
remaining audio 98. The resulting total audio signal output from
adder 99 is a combination of the preferred signal and remaining
audio where each signal has been independently gained and
"equalized."
[0126] Equalization refers to on method of frequency dependent
gain, similar to that which is used on many stereo systems. The
most elementary of equalizations modify the bass (low frequencies)
and treble (high frequencies) with a single knob. More
sophisticated equalizers have bandpass filters for every octave
band from 20 Hz to 20 kHz, each having a user adjustable gain.
[0127] The unique nature of this aspect of the present invention is
that the preferred signal can be equalized separately from all
other remaining audio, whereas conventional equalization only
modifies the total audio signal. One beneficial application of
equalization of a pure voice preferred signal can be realized by
raising the higher frequencies of the voice and lowering the high
frequencies of the remaining audio to improve speech
intelligibility, without raising the low frequency levels to a
point that is uncomfortable. The frequency dependent gain
components (96 and 98) can be realized in many possible ways, as
would be apparent to those skilled in the art, and are not
discussed in detail herein.
Adjustment in the Face of Transients
[0128] Most audio programming is transient in nature; that is, the
loudness increases or decreases readily over time. An analogy is
made to the programming of a football game where as the home team
approaches the goal line the crowds excitement grows as does the
loudness of their cheering. During the radio broadcast of such a
game, the crowd noise sometimes masks the announcer's voices.
However, this usually only lasts a few moments and the mixture of
crowd to announcer is back to the originally intended programming,
suitable for most listeners. The present invention described herein
assumes that the preferred signal (announcer's voice in this case)
is available as a separate component from the remaining audio
(crowd noise).
[0129] If this is the case, one additional aspect of the present
invention for correcting for transient changes in either the
preferred signal or remaining audio programming can maintain the
user selected gain ratio between the two signals throughout the
entire program, thus ensuring that the end user does not have to
continually adjust the two volume levels in the presence of
momentary swells in either signal. There are several ways to
accomplish this goal that are described next.
Soft Clipping Circuit
[0130] FIG. 11 illustrates one technique for automatically
maintaining the user selected ratio of preferred signal to
remaining audio, even in the presence of transient changes in
either program. Before the preferred signal or remaining audio is
delivered to the gain adjustment mechanisms (32, 33), each signal
is sent through a soft clipping circuit or algorithm 30, 31. As
before, the outputs of the gain adjustment mechanisms is added at
adder 34 and output to the end user as total
audio=w.sub.1PS+w.sub.2RA.
[0131] Soft clipping refers to the ability to limit the overall
magnitude of a signal when its level is too high, without
introducing any higher frequency artifacts. This is sometimes
referred to as companding as is well documented in its usage and
possible implementations. This technique implemented in the context
of automatically fixing the user adjusted PSRA ratio, can be
thought of as a two channel companding (or soft-clipping) circuit
with the ability to control the relative levels of the two limited
signals. The outputs of each of the two limiters (30, 31) can be
considered as normalized versions of their inputs based on some
preset parameter from the soft clipping circuit. Soft clipping can
be accomplished in a variety of ways with parameters that can be
programmed into the circuit or algorithm. It may also be
appropriate to offer the end user with the ability to adjust some
of these parameters to his/her own hearing needs.
Limiting Algorithm
[0132] Consider one exemplary embodiment of a limiting algorithm as
shown in FIG. 13. The input signal is collected over time and a
vector (string of input samples) is collected and temporarily
stored in storage 35. Once the vector has been filled with input
samples, it is delivered to an algorithm 36, which may determine
the peak signal value, the RMS of the vector, or some other
function of signal amplitude (which could also be a function of
frequency) that is then used to alter the input signal by simple
division by divider 37. The resulting signal is in a sense limited
by the amplitude of the original input signal.
[0133] There are at least two parameters in such an algorithm that
may be of interest for the user to control. The vector size 37 used
to generate the amplitude measure is a selectable quantity. For
large sizes the delay between the amplitude factor 36 and the
actual modified input signal may become quite large and cause the
effects of the limiting to be delayed and in some cases not
apparent. However, smaller sizes of the vector may not provide an
accurate representation for the total input signal, causing wide
variations in the amplitude factor 36 and resulting in wide
amplitude variations which may be unpleasant to listen to.
[0134] The method in which the amplitude factor is determined is
also a parameter that affects the end user's enjoyment of the
program. Selecting a maximum value of the vector will drastically
lower the overall level of the input signal in the presence of very
fast transient signals, and may lower the amplitude of other
signals in the audio that need to be heard immediately following
such a transient. Selecting the RMS level of the vector may also
not provide enough attenuation of the loud transients to make a
perceivable difference. It is likely that the choice of the peak
detection algorithm will be program and user dependent. So these
parameters may be preset or given as user selectable components for
both the preferred audio signal and the remaining audio signal.
[0135] Once the two signals have been limited, their relative
levels will remain constant even in the presence of transient
signals. Now the end user can apply separate gain (32, 33) to each
of the two limited signals to achieve the desired mixture between
the preferred signal and remaining audio. Transient surges in
either signal will be eliminated by the limiting circuit without
affecting the user selected adjustment of the relative signal
levels.
Additional Embodiment of Transient Suppression
[0136] Another method for accomplishing this same goal of
maintaining the user selected ratio of preferred signal to
remaining audio in the presence of transient signals, is presented
in FIG. 12. This method offers additional flexibility and will
support a more realistic reproduction of volume changes in either
the preferred signal or the remaining audio.
[0137] First, each signal is passed through its own volume
adjustment (38, 39) as selected by the user during any point in the
programming. Each signal is then amplified accordingly and sent to
vector accumulators (41, 42). The gain adjustment values (38, 39)
selected by the user are then used to generate a single number
ratio that represents the user selected preferred signal volume to
the user selected remaining audio volume (PSRA ratio).
[0138] Now, once the vector accumulators are full, an algorithm
similar to (36) realized as a peak detector, RMS calculator, etc.
(48, 49), computes the appropriate value that represents in some
form, a transient amplitude of the vector segment. The values for
each of the two signals are then used to make a ratio (43) that
represents the actual level of the preferred signal with respect to
the remaining audio after the user has adjusted the levels.
[0139] The user selected ratio and the actual ratio are compared by
division (44). If they are the same (as desired) the result is
unity and the preferred signal is not modified by the
multiplication operation (45). However, if the actual ratio is
greater than the desired ratio, the result of (44) will be less
than unity (perhaps very small). If the preferred signal is
multiplied by a very small number, the volume is greatly reduced
thus defeating the purpose of the entire user adjustment operation.
It is recognized that this may not be desirable in programs that
have very little or no remaining audio, such as news
programming.
[0140] One way to compensate for this problem is to ensure that the
result of (44) is never less than unity, thus only responding to
increases in the remaining audio volume. Another way to handle this
issue is to modify the remaining audio signal instead of the
preferred signal. Eliminating the multiplication (45) of the
preferred signal and adding an inverse operator (46) and
multiplication operation (47) to the remaining audio signal
accomplished this goal. Now, for increases in the transient signal
above the user desired ratio established with (38) and (39), the
system will decrease the amplitude of the remaining audio signal to
maintain the desired ratio. Again, precautions must be taken for
this configuration so that increases in the preferred signal do not
cause the transient level to become too loud. Also the parameters
for selecting vector size and determination of amplitude method may
be provided to the end user for a large amount of flexibility in
tailoring the audio programming.
Additional Embodiment for Transient Suppression
[0141] A final embodiment for automatically maintaining the user
selected ratio in the presence of transients in either the
preferred or remaining audio signals is seen in the form of
equalization. Coupling the previously described innovations for
frequency weighted gaining of the preferred signal separate from
the remaining audio signal with the automatic PSRA ratio hold
function, results in a system that can maintain different ratios of
preferred signal to remaining audio at different frequencies. So in
effect, the user selected frequency dependent gain may be at
different ratios for different frequency bands and the automatic
PSRA hold function will be operating separately at each of those
frequency bandwidths.
[0142] The implementation of this is realized simply as a bank of
systems such as the one shown in FIG. 12, that are each assigned to
a different frequency band, and each of the gains (38) and (39)
represent frequency dependent gains that operate only on a segment
of the total audio bandwidth for each of the preferred signal and
remaining audio signals.
Ambient Noise Adjustment
[0143] Another aspect of the present invention pertains to overall
programming levels for enjoyment in ambient noise environments. All
previous embodiments have focused on raising the preferred audio
level over the remaining audio which is present in the total audio
program that is initially delivered to the end user. This does not
address the environment where the end user is attempting to enjoy
the programming.
[0144] In steady state ambient noise fields, the user can simply
increase the volume of the programming to avoid masking that may be
occurring from noise present in his/her listening environment.
However, often transient noises require the user to continually
adjust the volume of the audio up and down in order to compensate
for such noises.
[0145] This aspect of the present invention automates the process
of manually adjusting audio programming to compensate for ambient
noises present in the listening environment. The process requires
that an independent measure of the ambient noise in the listening
environment be made simultaneously with the audio programming. The
reason this becomes a difficult prospect is because the audio
programming itself contributes to the overall ambient noise level
in the listening environment.
[0146] There are well established methods for extracting components
of signals given certain information about the system. The
measurement means used for determining the ambient noise in a
listening environment can take the form of any acousto-electric
transducer, such as a microphone. For the system described herein,
it will be assumed that the microphone is placed in front of the
electro-acoustic actuator means that is supplying the audio
programming to the listening environment.
[0147] This system is shown in FIG. 14. The speaker 151 represents
the electro-acoustic actuator which is placed near the microphone
152 measuring the ambient noise in the room.
[0148] The first step in obtaining a clean measurement of the
ambient noise, requires that the dynamics existing from the speaker
151 to the microphone 152 (all inclusive) be accurately identified
by system identifier 150. There are many algorithmic methods
available for system identification including the LMS algorithm
utilizing an FIR adaptive filter. System identification is a
requirement for the performance of the room audio monitoring system
to be effective. Once the physical system has been approximated
digitally, it is stored as a filter, G.sub.est.
[0149] FIG. 15 illustrates the entire system for adjusting the
volume of the audio programming in response to ambient noise
present in the listening environment. The block 158 represents the
ambient noise source in the listening environment monitored by the
microphone 157. First, the audio programming is filtered through
G.sub.est 153 to produce the approximate audio signal contribution
from the audio programming only, which is measured by the
microphone 157. The microphone measurement includes both the
ambient noise and the audio signal which is generated from the
speaker and travels through the air to the microphone 157.
[0150] To obtain a clean measure of the ambient noise in the
environment, the components of the audio signal generated by the
speaker are removed by adder 154. This only represents an
approximation of the ambient noise signal and is limited by the
accuracy of the system identification. Now, the approximation of
the audio programming components that are reaching the microphone
157 and thus the end user, are divided by the approximation of the
ambient noise levels by divider 155 to achieve a ratio which
represents an approximation of the actual volume of the audio
programming with respect to the ambient noise measured in the
listening environment.
[0151] Initially, the user has selected an overall volume of the
entire audio program (which could also be the volume of either the
preferred signal, the remaining audio, or the total audio volume
control) which is the desired volume level with respect to the
ambient noise in the room at the time of initial adjustment. This
value is then divided by the approximation of the actual volume
ratio of the audio programming to the ambient noise in the room to
resulting a single number which represents the difference in the
desired volume level with the approximated actual volume level.
This ratio is then used to control the mixer 159, which outputs the
actual audio signal until the desired ratio matches the actual
ratio. It is advisable that this algorithm be performed as an
average over time, such as the vector elements presented earlier.
This will make the changes in volume more pleasant and gradual. The
length of time that the vector covers may also become a user
adjustable parameter.
[0152] There are no limitations on the use of this embodiment of
the present invention in conjunction with the previously described
embodiments. In fact, a completely user adjustable audio system
will likely include each of these components used in conjunction
with each other.
User Control of Companding for Each Signal
[0153] The previous paragraphs discussed in detail several methods
for maintaining the user selected VRA in the face of transients in
either the voice or the remaining audio signal. Another key idea
related to the idea of maintaining the user selected VRA in the
face of audio transients, is that of allowing the user to adjust
the amount of companding (soft clipping, or limiting) that occurs
on each signal. This particular adjustment mechanism will allow the
user to maintain a specific improvement in the voice to remaining
audio adjustment while also experiencing a varying amount of
dynamic range adjustment on either (or both) of the voice or
remaining audio signals.
[0154] FIG. 16 illustrates this concept in terms of the amplitudes
of the time trace signals for voice 161 and remaining audio 162,
termed the dynamic amplitude range (DAR). Suppose the user adjusts
the voice level 161 to a mean value of 80 dB while the mean value
of the remaining audio is at 60 dB 162. Three different levels of
companding are shown from left to right for each signal since
either or both signals can experience compression.
[0155] If a high amount of compression is applied to both signals
as in 163, the DAR of both the voice and the remaining audio (loud
louds and soft softs) will be very compact. To achieve a more
realistic response to variations in amplitude response the listener
can adjust the DAR levels of both signals as in 164 or 165.
[0156] In the last case 165, it is clear that some of the remaining
audio amplitude shown by the dotted line will interfere with the
high amplitude voices. This case may happen when someone is trying
to yell over a gunshot or loud train. Clearly this provides a more
realistic performance but may be objectionable for those hard of
hearing.
[0157] As each signal's DAR can be adjusted independently, the user
may also select a narrow DAR for the voice along with a wide DAR
for the remaining audio, while the mean levels (also set by the
user) are maintained throughout the program.
Additional Embodiment for Transient Suppression
[0158] Another method for accomplishing the goal of maintaining the
user selected VRA ratio through transients, is to automatically
adjust the mean signal power levels (SPL) levels of the signals
when the remaining audio or voice becomes too loud to maintain the
user selected ratio.
[0159] Referring to FIG. 17, first consider an approximate moving
average of voice (solid) 171 and background audio (dotted) 172. The
average volume of the remaining audio frequently exceeds the
average volume of the voice, making dialog difficult to understand.
In an effort to achieve some improvement in intelligibility, the
user selects a VRA during a passage when the voice is fairly loud
with respect to the remaining audio, and he thinks he's achieved
his goal.
[0160] FIG. 18 illustrates such an adjustment using the same two
signals, i.e., the preferred signal (or voice) 181 and the
remaining audio (or background) 182. Even though the VRA has been
corrected during some segments of the program, other sections of
the remaining audio are still too loud due to transients.
Additional lowering of the remaining audio (increasing VRA) will
cause the sections of the remaining audio that are soft to be
completely imperceivable in the presence of dialog (or even due to
the dynamic range of the D/A hardware).
[0161] To solve this problem, only sections of the remaining audio
(or voice) that are too loud (or too soft) will be attenuated (or
amplified) in order to maintain the VRA set by the user at any
given moment in the programming.
[0162] FIG. 19 shows the sections where such an auto-leveling
process would work on the same signals discussed above. Section 193
represents the time during which the voice was raised to maintain
the VRA at a minimum of what was requested initially by the user.
Section 194 illustrates the section where the remaining audio was
attenuated to maintain a minimum VRA. Either or both of the
adjustments can be automated to ensure that the user maintains (as
a minimum) the VRA set at any point in time.
[0163] Additionally, FIG. 19 illustrates how some average DAR is
retained in the signals. Recall that the average deviation of these
signals can be limited on both amplitude levels so that either or
neither signal has very limited DAR.
Associated Audio Mode VRA Adjustment
[0164] Recall that there are Main, Associated, and Multi-lingual
services provided for by the ATSC standard. There are two main
services (complete main (CM), and music and effects (ME)) that are
typically 5.1 channels, and many associated services (dialog (D),
hearing impaired (HI), etc.) that are typically 1 channel. The
multi-lingual services could be achieved by providing a
multi-channel music and effects service (that does not contain
dialog) in conjunction with a dialog (D) associated service that is
reproduced primarily on the center channel. These two audio
services would be played simultaneously to constitute an entire
program. The dialog service can then be changed by the end user to
adjust the language (French, Spanish, etc.) without affecting the
original music and effects. If the ME channel is transmitted as a
5.1 channel program, the dialog is sent as a single channel
reproduced on the center channel, only without any spatial
positioning capability.
[0165] Some producers may be bothered by this; therefore a second
method for transmitting is to provide both the ME and D as
2-channel stereo audio allowing L-R positioning of both the
remaining audio and dialog. Currently, it seems that the use of the
dialog associated service is NOT the manner in which multiple
languages are delivered to the end user. Therefore, these aspects
of the present invention described next are envisioned for the time
when the associated services are utilized. This will likely occur
on future digital television broadcasts before DVD, since space is
of less concern on DVD's.
[0166] These aspects of the present invention take advantage of the
dialogue mode (or any of the associated audio services) by
extracting the associated service from the main service before they
are mixed in the decoder. The digital audio decoder decodes the
bitstream that contains both the main and associated service. (The
user typically selects which, if any, associated service besides
dialog is desired, i.e., commentary, voice over, etc., and also
selects the language in which the program will be viewed. This
information is all contained in the AC-3 bitstream as
metadata).
[0167] Once the decoder has uncompressed the audio signals,
extracted the metadata, and read the user inputs mentioned above,
the hardware selects the appropriate audio bitstreams and mixes
them together to form a total audio program. These channels are
then either given to the user as 5.1 channels, or downmixed to
fewer channels for systems having fewer speakers. This aspect of
the present invention will take the main audio and associated audio
(dialog service, hearing impaired service, visual impaired service,
etc.) before they are mixed, and apply adjustable gains to each of
them. This adjustability will be directly available to the user who
can lower or raise the relative level of either the voice or
remaining audio. Any spatial information is retained that was
present just before the mix. In addition, since the adjustment is
performed before the downmix, the user who has fewer than 5.1
channels can still take full advantage of the adjustability, while
the user with 5.1 channels can adjust the relative level of the
dialog on the center channel with respect to other audio on the
center channel if necessary.
[0168] As the dialog associated service (and other associated
services) described in A/54 might provide the voice separate from
the remaining audio all the way to the end user. However, the way
decoders are designed, these two signals are mixed together in the
consumer electronic hardware (set-top box, DVD player, DTV, any DD
decoder) before the signals reach the audio system. Any overall
level or frequency dependent level adjustment should also be
considered as part of this invention for any of the associated
services. The dialog service offers a particular benefit to
international products since the dialog mode is the mode that is
intended to support the multi-language capability. In addition, the
dialog mode contains speech that is not pre-processed by the
producer (which may be objectionable to those of normal hearing).
Therefore, the dialog level adjustment capability will offer the
normal listener the ability to adjust the overall level whereas the
hearing impaired listener may have the opportunity to apply user
specific processing to the normal voice, which is unavailable when
using the hearing impaired mode as described in the ATSC document
A/54.
[0169] In addition, any VRA type adjustment done using both the
associated service and the main audio services described above can
be used in conjunction with the VRA-hold devices mentioned above.
This automated capability is new to any relative level adjustment,
including what is implied by the HI mode (i.e., there are no
provisions for maintaining any of the user adjustments in the
presence of transients of either signal.) Therefore, this aspect of
the present invention allows: 1) main and associated audio service
relative level adjustment, i.e., VRA (for those services that don't
already employ such capabilities); and 2) a VRA-HOLD device, in
such a unique way that it can be easily built into existing decoder
designs without compromising the bitstream or requiring any
additional metadata or bandwidth.
[0170] Block diagrams of the hardware and software components for
this aspect of the present invention are depicted in FIG. 20, which
shows how the D and ME channels are delivered to the end user
having: 5.1 speakers (output of 205), 2 speakers (output of 206),
or one speaker (output of 207).
[0171] FIG. 20 represents the way audio is currently designed to be
transmitted to the end user on DVD using the dialog mode and the
multi-language services. DVD players or digital television
broadcasts that have the capability to transmit multiple languages,
permit the user to select the desired language through metadata
adjustments. The decoding, metadata and mixing all takes place
inside the chip hardware, either as part of the decoder itself or
as an integrated circuit that is also part of the decoder chip or
as a separate component. (The former is the most likely and least
expensive). The user does not have access to the dialog signal in
the current design because it was not envisioned to be
necessary.
[0172] According to the present invention, the decoder components
that perform the adding, be slightly altered as shown in FIG. 21.
After the bitstream is decoded by decoder 211 and the appropriate
metadata is applied (including the user selectable language), the
normal mode is as appears in FIG. 20, where the associated and main
audio services are mixed together (output of 215).
[0173] The difference now is that before the mixing occurs, the
same two signals are also sent directly to two (or one that
controls both) variable gain elements 212, 213 that are user
controllable. The sum 214 of the adjusted dialog and remaining
audio are delivered and downmixed to any number of speakers as
before. The user will also have the ability (either through
metadata or a switch) to toggle between the adjusted mode and the
normal mode via switch 216.
[0174] Another unique aspect of the present invention is that it
ultimately reduces the required bandwidth for providing the end
user (hearing impaired or otherwise) the capability to adjust the
dialog signal (in any language) in such a way that intelligibility
is optimized for that particular user. Currently, the vision is to
transport a hearing impaired dialog signal that has been "processed
for improved intelligibility" (via A/54) via the HI mode, in
conjunction with the dialog mode (which is a separate associated
audio service) containing multiple languages.
[0175] If the hearing impaired listener wants to hear the enhanced
dialog in another language, it may not be possible unless yet
another HI mode language is sent. Therefore, the current embodiment
may require four different dialog services to satisfy users for two
languages (for example) including: normal dialog English, normal
dialog French, hearing impaired mode English, and hearing impaired
mode French.
[0176] The present invention reduces this required bandwidth by
half and offers more functionality to a wider range of users! By
sending only the dialog mode (French and English) and breaking the
two signals (as shown above in the "VRA mode") before they are
combined in the decoder, the ability to adjust the level and apply
any post processing for the hearing impaired is available. As the
processing can be done at the convenience of the end user, it can
be tailored to the most desirable processing for that individual,
and not what the producer thinks is required.
[0177] Finally, breaking the dialog out of the decoder before it is
combined with the main service allows the dialog to be transmitted
to a hearing impaired listener directly, without interfering with
the viewing of those with normal hearing. This would typically be
accomplished via headphones (or via infrared or other wireless
means for sending the signal directly to a hearing aid being worn
by a hearing impaired listener) where the processed dialog is sent
directly to the hearing impaired individual and the normal dialog
is sent to the normal hearing listener through the air or another
set of headphones.
[0178] Ultimately, this aspect of the present invention provides:
1) improved VRA adjustability and truly personalized hearing
enhancement for the hearing impaired; 2) broader international
audience via the dialog mode; 3) reduced bandwidth for transmission
of digital audio; and 4) a method for accomplishing this without
modifying the AC-3 bitstream, without adding additional metadata,
and without hindering or changing any component of the encoding or
decoding process as is stated in the standards today.
[0179] Within the present invention are many other possible
embodiments; only a few of which are mentioned here. This
adjustment capability should not be limited to strictly volume, but
can include frequency shaping as well. In addition, we've primarily
discussed using this adjustment feature with the dialog service.
This capability can be made available on any of the associated
services just as easily.
[0180] Moreover, the prior discussion has only focused on
adjustments in terms of Dolby digital audio. These same ideas can
be applied to any similar audio format, including but not limited
to: DTS, SDDS, THX, MPEG, Dolby E, etc.
Center Channel Voice Adjustment
[0181] Audiophiles with "high-end" equipment including
multi-channel amplifiers and six-speaker systems, currently have a
limited capability to adjust the volume on the center channel
independently of the audio on the other five speakers. Since many
movies have mostly dialog on the center channel and other sound
effects located on other channels, this limited adjustment
capability allows the user to raise the amplitude of the mostly
dialog audio so that it is more intelligible during sections with
loud sound effects. Currently, this limited adjustment has two
important shortcomings and this aspect of the present invention
provides solutions to those limitations. The shortcomings are: It
is an adjustment capability that is only available to the consumers
that have a DVD player and a six-speaker home theater system that
permits volume level adjustment of all speakers independently; It
is an adjustment that is will need to be continuously modified
during transients in voice (center channel) and remaining audio
(all other channels); and VRA adjustments that were acceptable
during one audio segment may not be good for another audio segment
if the remaining audio level increases too much or the dialog level
reduces too much.
[0182] It is a fact that a large majority of consumers do not and
will not have a home theater that permits this adjustment
capability, i.e., Dolby Digital decoder, six-channel variable gain
amplifier and multi-speaker system for many years. In addition, no
consumer will have the ability to ensure that the VRA ratio
selected will stay the same for an entire program, until using the
present invention.
[0183] FIG. 22 illustrates the intended spatial positioning setup
of a common home theater system. Although there are no written
rules for audio production in 5.1 spatial channels, there are
industry standards. One of these standards is to locate the
majority of dialog on the center channel 226. Likewise other sound
effects that require spatial positioning will be placed on any of
the other four speakers labeled L 221, R 222, Ls 223, and Rs 224
for left, right, left surround and right surround. In addition, to
avoid damage to midrange speakers, low frequency effects LFE 225
are placed on the 0.1 channel directed toward a subwoofer
speaker.
[0184] Digital audio compression allows the producer to provide the
user with a greater dynamic range for the audio that was possible
through analog transmission. This greater dynamic range causes most
dialog to sound too low in the presence of some very loud sound
effects. The following example provides an explanation. Suppose an
analog transmission (or recording) has the capability to transmit
dynamic range amplitudes up to 95 dB and dialog is typically
recorded at 80 dB. Loud segments of remaining audio may obscure the
dialog when that remaining audio reaches the upper limit while
someone is speaking. However, this situation is exacerbated when
digital audio compression allows a dynamic range up to 105 dB.
Clearly, the dialog will remain at the same level (80 dB) with
respect to other sounds, only now the loud remaining audio can be
more realistically reproduced in terms of its amplitude. User
complaints that dialog levels have been recorded too low on DVD's
are very common. In fact, the dialog IS at the proper level and is
more appropriate and realistic than what exists for analog
recordings with limited dynamic range.
[0185] Even for consumers who currently have properly calibrated
home theater systems, dialog is frequently masked by the loud
remaining audio sections in many DVD movies produced today. A small
group of consumers are able to find some improvement in
intelligibility by increasing the volume of the center channel
and/or decreasing the volume of all of the other channels. However,
this fixed adjustment is only acceptable for certain audio passages
and it disrupts the levels from the proper calibration. The speaker
levels are typically calibrated to produce certain SPL's in the
viewing location. This proper calibration ensures that the viewing
is as realistic as possible. Unfortunately this means that loud
sounds are reproduced very loud (see dynamic range discussion
above). During late night viewing, this may not be desirable.
However, any adjustment of the speaker levels will disrupt the
calibration. The following invention will remedy this situation for
both normal (calibrated) viewing and late night (VRA) viewing.
Automatic VRA Adjustment feature for center channel
[0186] Some gain of the center channel level or reduction of the
remaining speaker levels provides improvement in speech
intelligibility for those consumers that have a 5.1 channel audio
system that has that adjustment capability. Note that all consumers
do not have such a system, and the present invention allows all
consumers to have that capability.
[0187] FIG. 23 illustrates a system where the user has the option
to select the automatic VRA leveling feature or the calibrated
audio feature. The system is calibrated by moving the switch to the
lower position which would be considered the normal operating
position where all 5.1 decoder output channels go directly to the
5.1 speaker inputs via a power amplifier (which could be
integrated).
[0188] The decoder would then be calibrated so that the speaker
levels were appropriate for the home theater system. As mentioned
earlier these speaker levels may not be appropriate for nighttime
viewing.
[0189] Therefore a second option (upper switch position in FIG. 23)
is made available through this aspect of the present invention,
that allows the consumer to select a desired VRA ratio and have it
automatically maintained by adjusting the relative levels of the
center channel with respect to the levels of the other audio
channels.
[0190] During segments of the audio program that don't violate the
user selected VRA, the speakers reproduce audio in the original
calibrated format. The auto-leveling feature only "kicks-in" when
the remaining audio becomes too loud or the voice becomes too soft.
During these moments, the voice level can be raised, the remaining
audio can be lowered, or a combination of both. This is
accomplished by the "check actual VRA" block 232 in FIG. 23. If the
user selects to have the auto VRA hold feature enabled via switch
235, then the 5.1 channels levels are compared in the check actual
VRA block 232. If the average center level is at a sufficient ratio
to that of the other channels (which could all be reverse
calibrated to match room acoustics and predicted SPL at the viewing
location) then the normal calibrated level is reproduced through
the amplifier 236 via fast switch 237.
[0191] If the ratio is predicted to be objectionable then the fast
switch 237 will deliver the center channel to its own auto-level
adjustment and all other speakers to their own auto level
adjustment. As mentioned earlier, the previous section discussed
these adjustment possibilities in terms of FIGS. 16-19 in detail.
Any of those aspects of the present invention can also be applied
to the 5.1 center/other speaker level adjustments as shown in FIG.
23.
[0192] According to the present invention: 1) those auto VRA-HOLD
features are applied directly to the existing 5.1 audio channels;
2) the center level that is currently adjustable in home theaters
can be adjusted to a specific ratio with respect to the remaining
channels and maintained in the presence of transients; 3) the
calibrated levels are reproduced when the user selected VRA is not
violated and are auto leveled when it is, thereby reproducing the
audio in a more realistic manner, but still adapting to transient
changes by temporarily changing the calibration; and 4) allowing
the user to select the auto (or manual) VRA or the calibrated
system, thereby eliminating the need for recalibration after center
channel adjustment. [note to add variability]
[0193] Also note that although the levels are said to be
automatically adjusted, that feature can also be disabled to
provide a simple manual gain adjustment as shown in FIG. 23.
Center Channel Adjustment for Downmix to Non-center Channel Speaker
Arrangements
[0194] Many consumers do not have and will not have home theater
systems for quite some time. However, DVD players are becoming more
popular and digital television will be broadcast in the near
future. These digital audio formats will require the end user to
have a 5.1 channel decoder in order to listen to any broadcast
audio, however, they may not have the luxury of buying a fully
adjustable and calibrated home theater system with 5.1 audio
channels.
[0195] The next aspect of the present invention takes advantage of
the fact that producers will be delivering 5.1 channels of audio to
consumers who may not have full reproduction capability, while
still allowing them to adjust the voice to remaining audio VRA
ratio level. In addition, this aspect of the present invention is
enhanced by allowing the user to choose features that will maintain
or hold that ratio without having a six speaker adjustable
system.
[0196] This can be realized by means of an interfacing unit that
receives a DD bitstream from the output port of a DVD player, so
another similar device, and then provides a means for adjusting the
center channel level, according to a user-selected VRA, a
customized audio decoder, followed by a downmix of the signals to
stereo, four-channel, or any other speaker arrangement that does
not provide a center channel speaker.
[0197] FIG. 24 illustrates one conceptual diagram of how this
downmix would be implemented.
[0198] FIG. 25 shows that this capability could be built into the
primary audio decoder that is utilized by any DVD, DTV, or any
other digital audio playback device that is intended to decode and
distribute audio signals to the speakers. The downmixing for the
non-home theater audio systems provides a method for all users to
benefit from a selectable VRA. The adjusted dialog, is distributed
to the non-center channel speakers in such a way as to leave the
intended spatial positioning of the audio program as intact as
possible. However, the dialog level will simply be higher. Further
details are discussed next.
[0199] There are well-specified guidelines for downmixing 5.1 audio
channels (Dolby Digital) to 4 channels (Dolby Pro-Logic), to 2
channels (stereo), or to 1 channel (mono). The proper combinations
of the 5.1 channels at the proper ratios were selected to produce
the optimum spatial positioning for whichever reproduction system
the consumer has. The problem with the existing methods of
downmixing is that they are transparent to and not controllable by
the end user. This can present problems with intelligibility, given
the manner in which dynamic range is utilized in the newer 5.1
channel audio mixes.
[0200] As an example, consider a movie that has been produced in
5.1 channels having a segment where the remaining audio masks the
dialog making it difficult to understand. If the consumer has 6
speakers and a 6 channel adjustable gain amplifier, speech
intelligibility can be improved and maintained as discussed above.
However, the consumer that has only stereo reproduction will
receive a downmixed version of the 5.1 channels conforming to the
diagram shown in FIG. 26 (taken from the Dolby Digital Broadcast
Implementation Guidelines). In fact, the center channel level is
attenuated by an amount that is specified in the DD bitstream
(either -3, -4.5 or -6 dB). This will further reduce
intelligibility in segments containing loud remaining audio on the
other channels.
[0201] This aspect of the present invention circumvents the
downmixing process by placing adjustable gain on each of the
spatial channels before they are downmixed to the users'
reproduction apparatus.
[0202] FIG. 27 illustrates the user adjustable levels on each of
the decoded 5.1 channels. Typically, downmixing of the low
frequency effects (LFE) channel is not done to prevent saturation
of electronic components and reduced intelligibility. However, with
user adjustment available before the downmix occurs, it is possible
to include the LFE in the downmix in a ratio specified by the
user.
[0203] Permitting the user to adjust the level of each channel
(level adjusters 276a-g) allows consumers having any number of
reproduction speakers to take advantage of the voice level
adjustment previously only available to those people who had 5.1
reproduction channels.
[0204] As shown above, this apparatus can be used external to any
DD decoder 271 whether it is a standalone decoder, inside a DVD, or
inside a television, regardless of the number of reproduction
channels are in the home theater system. The user must simply
command the decoder 271 to deliver a DD (5.1) output and the
"interface box" will perform the adjustment and downmixing,
previously performed by the decoder.
[0205] FIG. 28 illustrates this interface box 282. It can take as
its input, the 5.1 decoded audio channels from any DD decoder,
apply independent gain to each channel, and downmix according to
the number of reproduction speakers the consumer has.
[0206] In addition, this aspect of the present invention can be
incorporated into any decoder by placing independent user
adjustable channel gains on each of the 5.1 channels before any
downmixing is performed. The current method is to downmix as
necessary and then apply gain. This cannot improve dialog
intelligibility because for any downmix situation, the center is
mixed into the other channel containing remaining audio.
[0207] It should also be noted that the automatic VRA-HOLD
mechanisms discussed previously will be very applicable to this
embodiment. Once the VRA is selected by adjusting each amplifier
gain, the VRA-HOLD feature should maintain that ratio prior to
downmixing. Since the ratio is selected while listening to any
downmixed reproduction apparatus, the scaling in the downmixing
circuits will be compensated for by additional center level
adjustment applied by the consumer. So, no additional compensation
is necessary as a result of the downmixing process itself.
[0208] It should also be noted that bandpass filtering of the
center channel before user-adjusted amplification and downmixing
will remove sounds lower in frequency than speech and sound higher
in frequency than speech (200 Hz to 4000 Hz for example) and may
improve intelligibility in some passages. It is also very likely
that the content removed for improved intelligibility on the center
channel, also exists on the left and right channels since they are
intended for reproducing music and effects that would otherwise be
outside the speech bandwidth anyway. This will ensure that no loss
in fidelity of remaining audio sounds occurs while also improving
speech intelligibility.
[0209] This aspect of the present invention: 1) allows the consumer
having any number of speakers to take advantage of the VRA ratio
adjustment presently available to those having 5.1 reproduction
speakers; 2) allows those same consumers to set a desired dialog
level on the center channel with respect to the remaining audio on
the other channels, and have that ratio remain the same for
transients through the VRA-HOLD feature; and 3) can be applied to
any output of a DD decoder without modifying the bitstream or
increasing required transmission bandwidth, i.e., it is hardware
independent.
Hearing Impaired Voice Processing
[0210] Another aspect of the present invention includes specific
processing of voice on digital audio decoders for the benefit of
the hearing impaired. Specific filtering or frequency weighting of
the audio signal is required. The user-adjustment of the filters
from a Graphical User Interface will greatly simplify these
adjustments.
[0211] Another aspect would be to separately modify the magnitude
of the frequency response for both the dialog channel (possibly
center or pure dialog depending on the method chosen) and the
remaining audio. This would essentially be a graphic equalizer on
each channel to achieve maximum adjustability. However, it could
also include some fixed shapes for magnitude modification such as
high frequency amplification (for hearing impaired) of dialog to
improve intelligibility while the high frequencies of the remaining
audio would be attenuated.
[0212] Other effective processing, including special spatial
positioning of the voice for improved intelligibility, may also be
possible. For example, dialog may be easier to understand by the
hearing impaired if it is presented in a stereo format rather than
the more common center channel mono format.
[0213] Clearly on screen programming (OSP) will be an option for
any of the above ideas since they can all be implemented either as
part of the decoder or as a separate box. Knobs, dials, buttons
etc. can all be used to accomplish these goals as well.
[0214] The manual controls can consist of 1 or two knobs to do the
tasks, one for VRA or two for VRA, one for range of VRA deviation
or two for range of VRA deviation.
Digital Multi-Channel Decoders
[0215] This aspect of the present invention disclosed herein
provides the end user of (primarily) digital multi-channel audio
decoders, the ability to universally adjust a specified foreground
audio signal volume level and spatial location with respect to a
specified background audio signal volume level and spatial
location.
[0216] For a foreground signal including only speech or dialogue
programming, this technique will provide the ability to
dramatically improve speech intelligibility. This is accomplished
by using multiple independent channels to deliver coded audio
signals which are, to some degree, specified in terms of audio
content. The result is a set of components necessary to provide the
end user with this specific adjustment capability while using a
pre-specified digital decoder technology.
[0217] Additions to and modifications of the digital decoding
process are described to provide this end user adjustment
capability. Several options are described which will require some
modification to the digitally encoded signals, as are options that
will require only minimal modification of the encoded signals.
[0218] An algorithm is presented for use with a digital decoder
that will provide the user with the necessary adjustment capability
to adequately alter the speech intelligibility for a wide range of
programming content and media delivery methods.
Background
[0219] Recent advances in digital audio coding technology have
initiated wide-ranging changes in the future of audio broadcasting
and recording techniques. The Dolby Digital (AC-3) standard has
been accepted for use in digital-television and is currently (1998)
one of the competing audio formats for DVD. AC-3, or a similar
multi-channel coding technology will likely be instituted for
nearly every aspect of digital audio recording and broadcasting
throughout the next decade. Many new features are also included in
the AC-3 standard which provide the consumer electronics end user
to control aspects of the audio programming that were previously
uncontrollable.
[0220] The goal of this aspect of the present invention is to
provide further features previously unavailable for end user
enjoyment, thereby improving upon and expanding features currently
employed in existing audio decoder technology.
[0221] The actual coding and decoding process has been well
documented by Dolby and Toshiba, for the AC-3 encoder/decoder
designs. Likewise, other competing multi-channel audio coding
formats have been well specified and proven to provide five or more
audio channels while conserving signal bandwidth and required
storage media. Since these coding technologies are well developed
and many have become the standard for audio transmission and
storage, it is not the specific goal of these inventions to alter
the actual coding and decoding process.
[0222] In this and following descriptions "coding" and "decoding"
implies the compression and decompression of the bit streams after
they have been converted to digital audio. It also includes the
process of including other non-audio digital information that
informs the decoder about the type of audio-programming and options
available.
[0223] While one embodiment of this invention seeks to add audio
and non-audio information bearing signals to the existing audio
coding signals, it will do this while conforming to the coding
standards set forth for the particular media. For example, if the
inventions are to be applied to digital television (DTV)
broadcasting, and the audio standard for DTV broadcasting is AC-3,
this aspect of the present invention will conform to the standards
set forth in the description of that format. Other embodiments do
not require any specific additional information about the coding
process itself, in order to implement the inventions.
[0224] Therefore, given any multi-channel audio format (notably
"digital" but not restricted from analog) the specific user
adjustment technology disclosed herein will be implemented in a
variety of formats. Information about the technology can be
incorporated into the coded bit stream. The user adjustment
technology can be accomplished without specific modification of the
encoded bit stream. Additionally, the technology can be implemented
in any multi-channel digitally coded (or analog) variety of coding
schemes, without loss of generality.
Embodiments of the Decoder
[0225] The primary thrust of this technology seeks to provide the
end user with the ability to universally adjust the audio volume
levels of "background" audio with respect to "foreground" audio.
Specific implementation for any broadcast or recorded audio may
have the foreground audio as a speaker's voice whereas the
background audio will be any remaining audio components that are
non-voiced such as sound effects in a television show. If the
background audio is too loud, it will interfere with the end user's
ability to discern speech components. This aspect of the present
invention seeks to apply this adjustment capability to digital
audio decoders, thereby making the technology easily accessible to
a wide range of audio consumers on a variety of media formats.
[0226] The following descriptions use the Dolby AC-3 audio coding
scheme as an example for a digital implementation of this
technology. However, many other audio coding strategies
(including-DTS, THX, SDDS, linear PCM, MPEG-1, MPEG-2, etc.) exist.
The inventions of the digital implementations of this technology
are not restricted to AC-3, but can be used in any multi-channel
digital audio format. So without loss of generality for use in
conjunction with other audio standards, the following are described
next using the existing and well-defined AC-3 format.
[0227] Annex C of the AC-3 format explains an alternative use of
the multi-channel surround audio signals to perform Karaoke using
an AC-3 decoder equipped with the ability to decode and recognize
the Karaoke encoding scheme. (Karaoke is a means for entertainment
whereby music from songs is played without the vocal content so
that amateur performers may sing in place of the professional
recording artists). The Annex C describes a Karaoke aware decoder
that is equipped to handle the incoming Karaoke coded multi-channel
content. After the multi-channel signals have been decoded and are
formed into separate digital audio signals (called L--left stereo
audio, R--right stereo audio, M--"guide melody", V1--"vocal track
one", and V2--"vocal track two"), the following algebraic
operations occur:
L.sub.K==L+a*V1+b*V2+c*M
C.sub.K=d*V1+e*V2+f*M
R.sub.K=R+g*V1+h*V2+i*M
The resulting signals (called L.sub.K, C.sub.K, and R.sub.K) are
the Karaoke output signals corresponding to the left, center and
right audio channels, respectively. The coefficients (a, b, c, d,
e, f, g, h, i) are used to adjust the level of the signals that
they modify. The actual values of these coefficients are left up to
specific implementation, and are discussed in Annex C for the
Karaoke implementation only. There are several options for
alternate inventions regarding this algorithm (set of equations)
that will provide the end user with distinctly different end
results.
[0228] As mentioned earlier, the desired result is to provide the
end user (audio consumer) with the ability to improve the speech
intelligibility of audio programming by either 1) raising the level
of the foreground (desirable or speech) audio components
independent of the background (sound effects or other secondary
audio programming) audio components OR 2) lowering the background
audio components independent of the foreground audio
components.
[0229] In its current embodiment, the existing algorithm cannot
accomplish either of these goals because explicit control is not
established over the background audio (L and R), and the foreground
audio cannot be raised with respect to the background audio. Other
limitations involving the specified content of each of the Karaoke
decoder signals further prevents the end user to have the ability
to universally adjust audio programming for the purpose of
improving speech intelligibility or listening enhancement by
allowing the end user to adjust the relative levels of the voice
and remaining audio (hereafter referred to as universal
voice-to-remaining-audio (UVRA) adjustment). (Note: "Voice" in
general is the desired foreground signal that contains the most
information contained in the audio recording or broadcast. However,
the term "voice" should not limit the types of signals that may be
considered for use as foreground material).
[0230] One aspect of the present invention provides UVRA adjustment
capability lies in redefining the existing functioning of the AC-3
Karaoke aware decoder. Since other multi-channel codec's may have
similar Karaoke style features, these modification can be made to
any similar device without departing from this inventions original
intent. Using the existing algorithm for creating the output
channels from the signals L, R, M, V1, and V2, one can specify
specific content of each channel to correspond to certain
components of audio programming (broadcast or recording) AND
provide unique adjustment capability of the coefficients a, b, c,
d, e, f, g, h and i, such that speech intelligibility can be
improved by user adjustment of said coefficients.
[0231] One possible method for accomplishing this goal is to first
have all of these coefficients universally adjustable (essentially
from 0 to infinity, where infinity corresponds to the physical
limitations of the decoder and audio hardware, such as the
D/A).
[0232] Second, it should be noted that L and R cannot be adjusted
in the original set of equations used to present this first
inventive embodiment. This places limitations on what the signal
content for each channel can be if the D/A has a fixed output range
with no auto-scaling (addressed next). The V1, V2, and M channels
are universally adjustable and thus can have any content the
provider desires, as long as the voice (foreground) is kept
separate from the background. For example, M may contain all of the
foreground (desirable, or voice) audio components while the V1 and
V2 may contain sound effects, background music, or laughter. For
these cases, L and R must either contain no audio signals, or audio
that is at a low enough level that the other audio components may
be adjusted to levels high enough to obscure the L and R audio so
that it is nearly imperceivable. This method allows the end user to
adjust a, b, c, d, e, f, g, h and i so that the voice level can be
fixed independent of ANY remaining audio programming, and
positioned at any of three speaker locations (L.sub.O, R.sub.O,
C.sub.O), where the left output, right output and center output are
defined by:
L.sub.O=L+a*V1+b*V2+c*M
C.sub.O=d*V1+e*V2+f*M
R.sub.O=R+g*V1+h*V2+i*M
[0233] Mentioned earlier was the concept of output scaling (or
auto-scaling) and its use in conjunction with unspecified (or
specified) content on the decoded (but not output) left and right
channels depicted in the above equation and descriptions. Several
brief examples presented for the L.sub.O signal only will
illustrate the proposed operation of auto-scaling, required for the
above procedure to work without specifying content of the L and R
channels.
[0234] To accomplish the UVRA goal with the above linear
combination of decoded signals, the user would need to raise the
coefficient on V1 ("a") to a sufficient level above L and/or M in
order to achieve the desired VRA ratio. Suppose first that L and M
contain identical content that is the entire program (foreground
(voice) and (background) remaining audio) in the ratio that the
provider intended. Now suppose the V1 channel contains the vocal
(dialogue) content synchronized with the original programming. If
the user desires to have a 6 dB VRA he/she would select "a" to be
4.0 since the remaining audio is represented twice by L and M:
L.sub.O=L+4.0*V1+1.0*M
20*log.sub.10(4.0/2.0)=6 dB. Now, it is assumed that "full scale"
for the D/A hardware is represented by 0 dB. If L, V1, and M are
each at full signal scale (0 dB, or 1.0 linear), the linear
magnitude of L.sub.O becomes
L.sub.O=1.0+4.0(1.0)+1.0(1.0)=6, or 15.5 dB.
[0235] This level will saturate the D/A since full scale is 0 dB.
If UVRA auto-scaling is employed by dividing L.sub.O by its peak
magnitude, L.sub.O will be divided by 6.0 resulting in the
following expression:
L.sub.O=0.1667*L+0.6667*V1+0.1667*M
[0236] Note that the VRA ratio is still the desired
6 dB=20*log.sub.10(0.6667/(0.1667+0.1667)),
but L.sub.O=0 dB and will not saturate the D/A. The user has
achieved the goal of UVRA adjustment. However, it is critical to
note that this example was prepared for a certain method of
auto-scaling and a certain content of each of the previously
decoded signals.
[0237] One more example that will further generalize the possible
content of each of the channels is now presented. Suppose L and M
do not contain similar content and cannot be played simultaneously.
(A more generalized scenario that relaxes the assumptions on the
signal contents). However, the V1 channel still contains only the
programming dialogue or foreground and the M channel is still the
remaining audio or background. Therefore, a user adjusted mix of
only M and V1 would result in the desired UVRA adjustment
capability. Even though there is no explicit adjustment of the
remaining "L" signal provided by the algorithm shown above we can
effectively change the coefficient by an innovative use of the
auto-scaling process. (Note: this example can be generalized to be
used with ANY two signals as long as one of them has a volume
controller and auto-scaling works as described herein).
[0238] In order to accomplish the HEC goal with the signals
described above, two things need to occur:
[0239] Raise the voice to a desired level (say 6 dB) over the
remaining audio.
[0240] Raise the combined voice (V1) and remaining audio (M) to a
higher level than the L signal which has no adjustment capability,
and a fixed coefficient of 1.0.
[0241] First, in order for the remaining audio on the "M" channel
to sufficiently obscure the left channel audio "L", let's suppose
the level of the M channel must exceed the L channel by 20 dB. (It
may be more or less depending on the type of programming on each
channel). Therefore the coefficient of the M channel ("c") must be
10.0 since the coefficient of the "L" channel is 1.0
(20*log.sub.10(10.0/1.0)=20 dB). Now for the voice signal to be 6
dB greater than the remaining audio, the coefficient "a" must be
22.0 because 20*log.sub.10(22.0/11.0)=6.0 dB, where 11.0 is the sum
of the M channel and the L channel power. The overall equation then
becomes
L.sub.O=L+22.0*V1+10.0*M
[0242] Since full scale was set at 1.0 for L.sub.O, it is clear
that the above summation must be scaled as before to avoid D/A
saturation. The result of scaling of L.sub.O by 33.0 is:
L.sub.O=0.0303*L+0.6667*V1+0.4545*M
[0243] Note that the summation of all signals is unity (0 dB), the
VRA is 6 dB, and the VRA ratio of the foreground plus background
content (M+V1) to the unknown content (L) is 20 dB. This explains
how the specified method of auto-scaling can be used to allow the
original set of algebraic relations to result in UVRA adjustment
without explicit control over the coefficient of L. This
establishes the ability to provide end user UVRA adjustment without
affecting the standardized method for linear combinations. The
additional innovative components include channel content
specification, universal parameter adjustment and output signal
auto-scaling.
[0244] The next embodiment involves the development of an entirely
new algorithm that can be used in conjunction with multi-channel
audio coder's to deliver UVRA adjustment capability to the end
user, while still retaining much of the imaging capabilities
inherent in such codec's. The primary goal is to have the
programming provider (recorder, broadcast, or otherwise), deliver a
voice only signal of any type of audio programming on one or more
of these available multiple channels while simultaneously but
independently delivering the remaining audio components on the
other channels.
[0245] Some standardization needs to be agreed upon if the UVRA
adjustment capability is to be constructed into a variety of
decoders and mediums. The following algorithm seeks to provide that
standard, or some form of such a standard. As will be seen, it
matters little as to which channel(s) the voice is actually
transmitted on, since all components are adjustable. (This provides
an interesting feature to such a decoding scheme in that the UVRA
capability can be ignored without disturbing the normal
programming). The goal of multi-channel audio is to provide the end
user with a surrounding sound that makes audio programming seem
more lifelike. It is desirable to provide the end user with the
ability to adjust the VRA without significantly interfering with
the sound quality. One possible way that this can be accomplished
is by using the following algorithm:
L.sub.O=a*L+b*C
C.sub.O=c*L+d*R+e*C
L.sub.SO=h*L.sub.S+i*C
R.sub.SO=j*R.sub.S+k*C
[0246] The decoded audio signals are specified by L, R, C, L.sub.S,
and R.sub.S. The total output signals of the decoder (subscripted
by "O"), are algebraic functions of the decoded audio signals. The
manner in which the above algorithm has been developed provides
maximum adjustability, UVRA adjustment for location and level, and
minimal interference in the surround image that is normally created
for regular multi-channel audio programming.
[0247] In this embodiment, the audio signal C must be taken to be
the foreground audio. Assuming the foreground audio is considered
to be voice or speech, the user can adjust the level and location
of the voice by changing b, e, g, i, and/or k with respect to the
other user adjustable coefficients. The left audio output will
contain the left signal, the right audio output will contain the
right signal, and same with subsequent surround signals and
outputs.
[0248] The primary difference here is that the center audio output
is now a combination of the original left, right, and foreground
signals. In practice it is not uncommon to have the center channel
consist of primarily left and right components, sounds that do not
require any type of imaging.
[0249] It is inevitable that some spatiality ability be lost in
this design because one of the 5 original audio channels must be
occupied by pure foreground in order for the user to have control
over the relative levels. It is also important to note that for
ideal performance of the UVRA system, L, R, L.sub.S, and R.sub.S
should contain only background audio components. This will allow
the user to adjust the levels with respect to each other completely
independently. However, it is possible for these signals to contain
components of the foreground signal as long as they are properly
synchronized with the foreground audio (on the center channel in
this description) to avoid echoes and/or spectral cancellation.
[0250] Other forms of these equations are possible by redefining
the specific content of each of the original decoded audio signals,
that will ultimately accomplish the identical goal of UVRA and
spatial positioning of background and foreground audio independent,
or nearly independent of each other. As long as adjustment
capabilities are provided on the foreground signal with respect to
and independent of the background signal, the goal of UVRA
adjustment (both volume and spatial positioning) will be
accomplished and the intent of this invention will be realized.
[0251] It should be clear that the above procedure can be carried
out without any specific modification to the decoder itself. If the
linear combination of the decoder signals is carried out within the
decoder hardware, it is obvious that modifications must also occur
to those particular algorithms.
[0252] However, it is not necessary if the program provider
supplies the encoder with the signals as mentioned above, and the
decoder provides the end user with the raw decoded audio signals
prior to linear combination. In this embodiment, a second piece of
hardware can be used to perform the linear combination of these
signals (which can be analog or digital) and provide the end user
with the specified adjustment capability.
[0253] Alternately, modification of the digitally encoded bit
stream may also be performed to inform the decoder that a special
"UVRA organized" signal is (optionally) available. This may permit
decoder to provide the end user with a choice between normal
multi-channel audio programming and a special UVRA adjustable
multi-channel signal. If the user selects the UVRA audio stream
instead of the normal audio stream, the adjustment capabilities can
be performed either on an external unit or internally to the
decoder itself prior to the D/A operation.
[0254] Other information can also be added to the bit stream by the
program provider informing the end user of the channel that the
foreground audio is on, the desired spatial arrangement, and
suggested coefficients for proper positioning or placement of the
foreground and background audio signals.
[0255] To summarize, this aspect of the present invention relies on
the ability to deliver the foreground audio separately from the
background audio. Many multi-channel audio formats delivery of such
signals although they do not specify content of each signal beyond
the spatiality. With this aspect of the present invention, for some
sacrifice in spatiality, the end user can receive the ability to
universally adjust the foreground audio with respect to the
background audio providing at least one obvious result that speech
intelligibility can be improved for the case where foreground audio
is voice and background audio is an (otherwise) noisy signal that
may frequently obscure the voice.
[0256] This aspect of the present invention can be provided in a
variety of audio codec formats in manners that may also involve
limited (including special scaling abilities) and/or unlimited
adjustment of channel coefficients, specified and/or loosely
specified content descriptions of channels, modifying the decoder
bit stream and/or not modifying the bit stream (with cooperation
from audio programming providers), and providing alternate but
useful information concerning the use of the UVRA signal grouping
or receiving normal multi-channel audio programming.
[0257] While many changes and modifications can be made to the
invention within the scope of the appended claims, such changes and
modifications are within the scope of the claims and covered
thereby.
* * * * *